diff --git a/.github/workflows/ci-daily.yml b/.github/workflows/ci-daily.yml
index 009e6c523..dea30ab7e 100644
--- a/.github/workflows/ci-daily.yml
+++ b/.github/workflows/ci-daily.yml
@@ -8,7 +8,8 @@ on:
 env:
   NPROC: 2
   MAKEFLAGS: "-j${NPROC}"
-  NIMFLAGS: "--parallelBuild:${NPROC} --colors:off -d:chronicles_colors:none"
+  NIMFLAGS: "--parallelBuild:${NPROC} --colors:off -d:chronicles_colors:none -d:disableMarchNative"
+  NIM_PARAMS: "-d:disableMarchNative"
 
 jobs:
   build:
diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index 84a7f0b8d..9ddf904ef 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -13,7 +13,8 @@ concurrency:
 env:
   NPROC: 2
   MAKEFLAGS: "-j${NPROC}"
-  NIMFLAGS: "--parallelBuild:${NPROC} --colors:off -d:chronicles_colors:none"
+  NIMFLAGS: "--parallelBuild:${NPROC} --colors:off -d:chronicles_colors:none -d:disableMarchNative"
+  NIM_PARAMS: "-d:disableMarchNative"
   NIM_VERSION: '2.2.4'
   NIMBLE_VERSION: '0.22.3'
 
@@ -160,7 +161,7 @@ jobs:
           fi
 
           export MAKEFLAGS="-j1"
-          export NIMFLAGS="--colors:off -d:chronicles_colors:none"
+          export NIMFLAGS="--colors:off -d:chronicles_colors:none -d:disableMarchNative"
           export USE_LIBBACKTRACE=0
 
           make V=1 POSTGRES=$postgres_enabled test
diff --git a/.github/workflows/version-check.yml b/.github/workflows/version-check.yml
index e3ad958ba..ee01a9f1a 100644
--- a/.github/workflows/version-check.yml
+++ b/.github/workflows/version-check.yml
@@ -28,8 +28,11 @@ jobs:
           set -euo pipefail
           NIMBLE_VERSION=$(grep -m1 '^version = ' waku.nimble | sed -E 's/version = "([^"]+)"/\1/')
           # Nearest tag reachable from HEAD; --abbrev=0 drops the -<n>-g<sha>
-          # suffix so we get the bare tag (e.g. v0.38.0).
-          BASE_TAG=$(git describe --tags --abbrev=0 2>/dev/null || echo "")
+          # suffix so we get the bare tag (e.g. v0.38.0). `--match 'v*'` skips
+          # the moving `nightly` tag (auto-updated by the daily CI to point at
+          # master HEAD), which would otherwise be picked as the nearest tag
+          # and break the version-sort comparison below.
+          BASE_TAG=$(git describe --tags --abbrev=0 --match 'v*' 2>/dev/null || echo "")
           BASE_TAG=${BASE_TAG#v}
           # Compare on the base version, ignoring any -rc.N prerelease suffix.
           BASE_TAG=${BASE_TAG%%-*}
diff --git a/channels/events.nim b/channels/events.nim
index 2ee20a26d..3e271976e 100644
--- a/channels/events.nim
+++ b/channels/events.nim
@@ -21,3 +21,19 @@ EventBroker:
     channelId*: ChannelId
     senderId*: SdsParticipantID
     payload*: seq[byte]
+
+EventBroker:
+  ## Emitted when every segment of a channel-level `send()` reached
+  ## `Confirmed`. Channel-level analogue of `MessageSentEvent`; the
+  ## `requestId` is the channel-layer parent returned by `send()`.
+  type ChannelMessageSentEvent* = object
+    channelId*: ChannelId
+    requestId*: RequestId
+
+EventBroker:
+  ## Emitted when a channel-level `send()` finalises with at least one
+  ## segment in `Failed`. Channel-level analogue of `MessageErrorEvent`.
+  type ChannelMessageErrorEvent* = object
+    channelId*: ChannelId
+    requestId*: RequestId
+    error*: string
diff --git a/channels/rate_limit_manager/rate_limit_manager.nim b/channels/rate_limit_manager/rate_limit_manager.nim
index 5ea6486a5..ab5a9f67b 100644
--- a/channels/rate_limit_manager/rate_limit_manager.nim
+++ b/channels/rate_limit_manager/rate_limit_manager.nim
@@ -29,7 +29,7 @@ EventBroker:
   ##
   ## `channelId` lets listeners filter to their own channel, since all
   ## reliable channels share the underlying Waku node's broker context.
-  type ReadyToSendEvent* = object
+  type ReadyToSendEvent* = ref object
     channelId*: SdsChannelID
     msgs*: seq[seq[byte]]
 
diff --git a/channels/reliable_channel.nim b/channels/reliable_channel.nim
index f6a1f2ec4..23435030e 100644
--- a/channels/reliable_channel.nim
+++ b/channels/reliable_channel.nim
@@ -13,14 +13,14 @@
 ##
 ## See: https://lip.logos.co/messaging/raw/reliable-channel-api.html
 
-import std/[options, tables]
+import std/[options, sets, tables]
 import results
 import chronos
 import bearssl/rand
 import stew/byteutils
 import libp2p/crypto/crypto as libp2p_crypto
 
-import waku/messaging_client
+import waku/api/types
 import waku/node/delivery_service/send_service
 import waku/waku_core/topics
 
@@ -31,8 +31,8 @@ import ./rate_limit_manager/rate_limit_manager
 import ./encryption/encryption
 
 export
-  messaging_client, send_service, events, segmentation, scalable_data_sync,
-  rate_limit_manager, encryption
+  types, send_service, events, segmentation, scalable_data_sync, rate_limit_manager,
+  encryption
 
 const LipWireReliableChannelVersion* = "RELIABLE-CHANNEL-API/1"
   ## Wire-format spec marker for the Reliable Channel layer, as defined
@@ -42,27 +42,65 @@ const LipWireReliableChannelVersion* = "RELIABLE-CHANNEL-API/1"
   ## The trailing `/N` is the wire-format version and is bumped only
   ## on breaking on-the-wire changes; implementations pin one version.
 
-type ReliableChannel* = ref object
-  ## Spec-defined public type. Fields are private so callers cannot
-  ## mutate internals and break invariants. Getters are added below
-  ## for the few values consumers may need.
-  messagingClient: MessagingClient
-  channelId: ChannelId
-  contentTopic: ContentTopic
-  senderId: SdsParticipantID
-  rng: ref HmacDrbgContext
-  segmentation: SegmentationHandler
-  sdsHandler: SdsHandler
-  rateLimit: RateLimitManager
+type
+  SendHandler* = proc(envelope: MessageEnvelope): Future[Result[RequestId, string]] {.
+    async: (raises: [CatchableError]), gcsafe
+  .}
+    ## Egress dispatch boundary. Typically wraps `MessagingClient.send`;
+    ## tests inject a fake that records calls and returns canned
+    ## `RequestId`s so the send state machine can be exercised end-to-end
+    ## without a network.
 
-  requestIds: Table[RequestId, seq[RequestId]]
-  pendingRequests: seq[tuple[parent: RequestId, ephemeral: bool]]
-  brokerCtx: BrokerContext
-    ## Captured here so the channel emits `ChannelMessageReceivedEvent`
-    ## on the same broker context the owning manager registered its
-    ## listeners on. Without this, an emit via `globalBrokerContext()`
-    ## would land on whatever context happens to be thread-local at
-    ## emit time, which is not necessarily the manager's.
+  MessagePersistence {.pure.} = enum
+    Persistent
+    Ephemeral
+
+  SegmentSendState {.pure.} = enum
+    ## Lifecycle of a single segment as tracked by the channel. The
+    ## messaging layer has its own richer `DeliveryState` (retries,
+    ## propagated-vs-validated); here we only model what's needed to
+    ## decide when a `channelReqId` is fully accounted for.
+    AwaitingRateLimit ## Pushed by `send`; not yet released by rate_limit_manager.
+    InFlight
+      ## Released by rate_limit_manager and handed to delivery_service;
+      ## `messagingReqId` is now set.
+    Confirmed ## `MessageSentEvent` arrived for `messagingReqId`.
+    Failed
+      ## `MessageErrorEvent` arrived for `messagingReqId`, or the local
+      ## delivery-task construction failed before any id was reachable.
+
+  PendingMessagingRequest = object
+    ## One entry per segment (i.e. per messaging-layer request). The
+    ## relative order of `AwaitingRateLimit` entries must match the
+    ## order in which `rate_limit_manager` re-emits messages, which is
+    ## FIFO with `send()`.
+    channelReqId*: RequestId
+      ## The channel-layer parent id returned to the caller of `send()` in channel layer.
+      ## One channel request maps to N pending messaging requests.
+    messagingReqId*: Option[RequestId]
+      ## Per-segment messaging layer id. `none` until `onReadyToSend` assigns it.
+    persistenceReqType: MessagePersistence
+    segmentSendState*: SegmentSendState
+
+  ReliableChannel* = ref object
+    ## Spec-defined public type. Fields are private so callers cannot
+    ## mutate internals and break invariants. Getters are added below
+    ## for the few values consumers may need.
+    sendHandler: SendHandler
+    channelId: ChannelId
+    contentTopic: ContentTopic
+    senderId: SdsParticipantID
+    rng: ref HmacDrbgContext
+    segmentation: SegmentationHandler
+    sdsHandler: SdsHandler
+    rateLimit: RateLimitManager
+
+    requestIds: Table[RequestId, seq[RequestId]]
+    pendingMessagingRequests: seq[PendingMessagingRequest]
+      ## Entries are kept until the matching segment reaches a final
+      ## state (`Confirmed` or `Failed`); a whole channel request is
+      ## then pruned in one pass once all its segments are final.
+    brokerCtx: BrokerContext
 
 func getChannelId*(self: ReliableChannel): ChannelId {.inline.} =
   self.channelId
@@ -73,19 +111,103 @@ func getContentTopic*(self: ReliableChannel): ContentTopic {.inline.} =
 func getSenderId*(self: ReliableChannel): SdsParticipantID {.inline.} =
   self.senderId
 
+func isFinal(state: SegmentSendState): bool {.inline.} =
+  return state in {SegmentSendState.Confirmed, SegmentSendState.Failed}
+
+proc pruneCompletedChannelReqs(self: ReliableChannel) =
+  ## Drop every `pendingMessagingRequests` entry whose `channelReqId`
+  ## has all of its segments in a final state. A single failing
+  ## segment doesn't trigger a drop on its own — we wait until siblings
+  ## are also accounted for, so the channel-level outcome is decided
+  ## from a complete picture. For each fully-final `channelReqId`, emit
+  ## the channel-level final event before the entries are dropped:
+  ## `ChannelMessageSentEvent` if every sibling Confirmed,
+  ## `ChannelMessageErrorEvent` if any sibling Failed.
+  var hasPending = initHashSet[RequestId]()
+  var anyFailed = initHashSet[RequestId]()
+  for entry in self.pendingMessagingRequests:
+    if not entry.segmentSendState.isFinal():
+      hasPending.incl(entry.channelReqId)
+    elif entry.segmentSendState == SegmentSendState.Failed:
+      anyFailed.incl(entry.channelReqId)
+
+  var emitted = initHashSet[RequestId]()
+  for entry in self.pendingMessagingRequests:
+    if entry.channelReqId in hasPending or entry.channelReqId in emitted:
+      continue
+    emitted.incl(entry.channelReqId)
+    if entry.channelReqId in anyFailed:
+      ChannelMessageErrorEvent.emit(
+        self.brokerCtx,
+        ChannelMessageErrorEvent(
+          channelId: self.channelId,
+          requestId: entry.channelReqId,
+          error: "one or more segments failed",
+        ),
+      )
+    else:
+      ChannelMessageSentEvent.emit(
+        self.brokerCtx,
+        ChannelMessageSentEvent(
+          channelId: self.channelId, requestId: entry.channelReqId
+        ),
+      )
+
+  self.pendingMessagingRequests.keepItIf(it.channelReqId in hasPending)
+
+proc onMessageSent(self: ReliableChannel, messagingReqId: RequestId) =
+  ## Invoked from this channel's `MessageSentEvent` listener. Flips
+  ## the matching `InFlight` segment to `Confirmed` and prunes. The
+  ## listener routes every event through here; entries that don't
+  ## belong to this channel simply don't match and are no-ops.
+  self.pendingMessagingRequests.applyItIf(
+    it.segmentSendState == SegmentSendState.InFlight and
+      it.messagingReqId == some(messagingReqId)
+  ):
+    it.segmentSendState = SegmentSendState.Confirmed
+  self.pruneCompletedChannelReqs()
+
+proc onMessageError(self: ReliableChannel, messagingReqId: RequestId) =
+  ## Symmetric to `onMessageSent` but for `MessageErrorEvent`.
+  self.pendingMessagingRequests.applyItIf(
+    it.segmentSendState == SegmentSendState.InFlight and
+      it.messagingReqId == some(messagingReqId)
+  ):
+    it.segmentSendState = SegmentSendState.Failed
+  self.pruneCompletedChannelReqs()
+
 proc onReadyToSend(
-    self: ReliableChannel, msgs: seq[seq[byte]]
+    self: ReliableChannel, readyToSendEvent: ReadyToSendEvent
 ) {.async: (raises: []).} =
   ## Tail of the outgoing pipeline. Invoked from the `ReadyToSendEvent`
   ## listener once `rate_limit_manager` releases a batch of opaque
   ## blobs (already-encoded SDS messages):
   ##
   ##   ... -> rate_limit_manager -> [encryption] -> dispatch
-  for m in msgs:
-    ## Each `m` was preceded by exactly one push onto `pendingRequests`
-    ## in `send`, so this pop is always safe in the current skeleton.
-    let pending = self.pendingRequests[0]
-    self.pendingRequests.delete(0)
+  var idx = 0
+  for m in readyToSendEvent.msgs:
+    ## The first `AwaitingRateLimit` entry in push order is the one
+    ## this `m` belongs to: `send()` adds one entry per segment, and
+    ## `rate_limit_manager` re-emits them in the same FIFO order, so
+    ## the two sequences advance in lockstep. Earlier entries may
+    ## already be `InFlight` / `Confirmed` / `Failed` because they
+    ## live on until every sibling of their `channelReqId` is final,
+    ## so we walk past those to find the next one that was awaiting for this batch.
+    while idx < self.pendingMessagingRequests.len and
+        self.pendingMessagingRequests[idx].segmentSendState !=
+        SegmentSendState.AwaitingRateLimit
+    :
+      idx.inc()
+    if idx >= self.pendingMessagingRequests.len:
+      ## rate_limit_manager emitted more messages than we have pending —
+      ## should not happen given `send` pushes one entry per enqueued
+      ## SDS payload. Drop silently rather than corrupt state.
+      break
+
+    let channelReqId = self.pendingMessagingRequests[idx].channelReqId
+    let isEphemeral =
+      self.pendingMessagingRequests[idx].persistenceReqType ==
+      MessagePersistence.Ephemeral
 
     ## TODO: revisit which fields of the SDS message must be encrypted.
     ## Encrypting the whole encoded blob forces every receiver to attempt
@@ -97,32 +219,58 @@ proc onReadyToSend(
       MessageErrorEvent.emit(
         self.brokerCtx,
         MessageErrorEvent(
-          requestId: pending.parent,
-          messageHash: "",
-          error: "encryption failed: " & error,
+          requestId: channelReqId, messageHash: "", error: "encryption failed: " & error
         ),
       )
+      ## Encryption failed *before* we could hand the segment to the
+      ## delivery layer — no `messagingReqId` was minted and no
+      ## `DeliveryTask` was queued on `sendService`. The delivery
+      ## layer will therefore never emit a `MessageSentEvent` /
+      ## `MessageErrorEvent` for this segment, so `onMessageError`
+      ## won't fire either. Advance the state machine inline so the
+      ## parent `channelReqId` can still be pruned once its siblings
+      ## are also final.
+      self.pendingMessagingRequests[idx].segmentSendState = SegmentSendState.Failed
+      idx.inc()
       continue
     let wireBytes = seq[byte](encrypted)
 
+    ## The `meta` field carries the Reliable Channel wire-format spec
+    ## marker so the ingress side of any peer can route this WakuMessage
+    ## to its Reliable Channel layer.
     let envelope = MessageEnvelope(
-      contentTopic: self.contentTopic, payload: wireBytes, ephemeral: pending.ephemeral
+      contentTopic: self.contentTopic,
+      payload: wireBytes,
+      ephemeral: isEphemeral,
+      meta: LipWireReliableChannelVersion.toBytes(),
     )
 
-    let deliveryReqId = RequestId.new(self.rng)
-    let deliveryTask = DeliveryTask.new(deliveryReqId, envelope, globalBrokerContext()).valueOr:
-      ## TODO: emit waku `MessageErrorEvent` for the parent request id.
+    ## `sendHandler` is not annotated `(raises: [])`, but this listener is.
+    ## Convert any raise to a Result error so the state machine handles
+    ## both failure modes (Result.err and exception) through one path.
+    let sendRes =
+      try:
+        await self.sendHandler(envelope)
+      except CatchableError as e:
+        Result[RequestId, string].err("messaging send raised: " & e.msg)
+
+    let messagingReqId = sendRes.valueOr:
+      MessageErrorEvent.emit(
+        self.brokerCtx,
+        MessageErrorEvent(
+          requestId: channelReqId, messageHash: "", error: "messaging send failed: " & error
+        ),
+      )
+      self.pendingMessagingRequests[idx].segmentSendState = SegmentSendState.Failed
+      idx.inc()
       continue
 
-    ## Stamp the Reliable Channel wire-format spec marker so the ingress
-    ## side of any peer can route this WakuMessage to its Reliable
-    ## Channel layer. Done on the constructed WakuMessage rather than
-    ## via the envelope because `MessageEnvelope` does not expose a
-    ## `meta` field.
-    deliveryTask.msg.meta = LipWireReliableChannelVersion.toBytes()
+    self.pendingMessagingRequests[idx].messagingReqId = some(messagingReqId)
+    self.pendingMessagingRequests[idx].segmentSendState = SegmentSendState.InFlight
+    self.requestIds.mgetOrPut(channelReqId, @[]).add(messagingReqId)
+    idx.inc()
 
-    asyncSpawn self.messagingClient.sendService.send(deliveryTask)
-    self.requestIds.mgetOrPut(pending.parent, @[]).add(deliveryReqId)
+  self.pruneCompletedChannelReqs()
 
 proc send*(
     self: ReliableChannel, payload: seq[byte], ephemeral: bool = false
@@ -135,18 +283,22 @@ proc send*(
   ##
   ## `rate_limit_manager.enqueueToSend` emits a `ReadyToSendEvent` with
   ## the SDS messages cleared for transmission; the channel's listener
-  ## then runs the final stage (encryption -> dispatch). The `ephemeral`
-  ## flag is carried alongside each segment in `pendingRequests` and
-  ## stamped onto the eventual `MessageEnvelope`.
+  ## then runs the final stage (encryption -> dispatch). The
+  ## `persistenceReqType` is carried alongside each segment in
+  ## `pendingMessagingRequests` and stamped onto the eventual
+  ## `MessageEnvelope`.
   ##
-  ## The returned `RequestId` is the parent of one-or-more
-  ## delivery-service `RequestId`s; the mapping is recorded in
+  ## The returned `RequestId` is the channel-level parent of one-or-more
+  ## messaging-layer `RequestId`s; the mapping is recorded in
   ## `self.requestIds`.
   if payload.len == 0:
     return err("empty payload")
 
-  let parentReqId = RequestId.new(self.rng)
-  self.requestIds[parentReqId] = @[]
+  let channelReqId = RequestId.new(self.rng)
+  self.requestIds[channelReqId] = @[]
+
+  let persistenceReqType =
+    if ephemeral: MessagePersistence.Ephemeral else: MessagePersistence.Persistent
 
   for segmentBytes in self.segmentation.performSegmentation(payload):
     ## Segments arrive already encoded; the segmentation module owns
@@ -155,10 +307,17 @@ proc send*(
       self.channelId, self.senderId, segmentBytes
     ).valueOr:
       return err("SDS wrap failed: " & error)
-    self.pendingRequests.add((parent: parentReqId, ephemeral: ephemeral))
+    self.pendingMessagingRequests.add(
+      PendingMessagingRequest(
+        channelReqId: channelReqId,
+        messagingReqId: none(RequestId),
+        persistenceReqType: persistenceReqType,
+        segmentSendState: SegmentSendState.AwaitingRateLimit,
+      )
+    )
     self.rateLimit.enqueueToSend(sdsBytes)
 
-  return ok(parentReqId)
+  return ok(channelReqId)
 
 proc onMessageReceived(
     self: ReliableChannel, messageHash: string, payload: seq[byte]
@@ -206,7 +365,7 @@ proc onMessageReceived(
 
 proc new*(
     T: type ReliableChannel,
-    messagingClient: MessagingClient,
+    sendHandler: SendHandler,
     channelId: ChannelId,
     contentTopic: ContentTopic,
     senderId: SdsParticipantID,
@@ -221,8 +380,12 @@ proc new*(
   ## should be wiring up. Encryption is delegated to the `Encrypt`/
   ## `Decrypt` request brokers, so the channel keeps no per-instance
   ## encryption state either.
+  ##
+  ## `sendHandler` is the egress dispatch. The owning `ReliableChannelManager`
+  ## typically constructs it as a closure over `MessagingClient.send`. Tests
+  ## pass a fake to drive the send state machine without touching the network.
   let chn = T(
-    messagingClient: messagingClient,
+    sendHandler: sendHandler,
     channelId: channelId,
     contentTopic: contentTopic,
     senderId: senderId,
@@ -231,20 +394,21 @@ proc new*(
     sdsHandler: SdsHandler.new(sdsConfig, senderId),
     rateLimit: RateLimitManager.new(rateConfig, channelId, brokerCtx),
     requestIds: initTable[RequestId, seq[RequestId]](),
-    pendingRequests: @[],
+    pendingMessagingRequests: @[],
     brokerCtx: brokerCtx,
   )
 
-  ## Each channel owns its own egress + ingress listeners on
-  ## `chn.brokerCtx`, filtered to traffic addressed to this channel.
-  ## Keeping the listeners (and the procs they call) inside the
-  ## channel lets `onReadyToSend` and `onMessageReceived` stay private
-  ## — the manager doesn't need to know about them.
+  ## Each channel owns its own egress + ingress + send-completion
+  ## listeners on `chn.brokerCtx`, filtered to traffic addressed to
+  ## this channel. Keeping the listeners (and the handler procs they
+  ## call) inside the channel lets `onReadyToSend` /
+  ## `onMessageReceived` / `onMessageSent` / `onMessageError` stay
+  ## private — the manager doesn't need to know about them.
   discard ReadyToSendEvent.listen(
     chn.brokerCtx,
     proc(evt: ReadyToSendEvent): Future[void] {.async: (raises: []).} =
       if evt.channelId == chn.channelId:
-        await chn.onReadyToSend(evt.msgs)
+        await chn.onReadyToSend(evt)
     ,
   )
 
@@ -261,4 +425,20 @@ proc new*(
     ,
   )
 
+  ## Send-completion events are tagged with the per-segment messaging
+  ## `requestId` — globally unique, so we don't need any channel filter
+  ## up front. The handler scans this channel's pending entries for a
+  ## match and is a no-op when the id belongs to a different channel.
+  discard MessageSentEvent.listen(
+    chn.brokerCtx,
+    proc(evt: MessageSentEvent): Future[void] {.async: (raises: []).} =
+      chn.onMessageSent(evt.requestId),
+  )
+
+  discard MessageErrorEvent.listen(
+    chn.brokerCtx,
+    proc(evt: MessageErrorEvent): Future[void] {.async: (raises: []).} =
+      chn.onMessageError(evt.requestId),
+  )
+
   return chn
diff --git a/channels/reliable_channel_manager.nim b/channels/reliable_channel_manager.nim
index c94baa836..020bf9b2d 100644
--- a/channels/reliable_channel_manager.nim
+++ b/channels/reliable_channel_manager.nim
@@ -24,23 +24,28 @@ export reliable_channel
 type ReliableChannelManager* = ref object
   channels: Table[ChannelId, ReliableChannel]
   messagingClient: MessagingClient
-    ## Borrowed from the owning `Waku`. The ownership chain is
-    ##   Waku -> ReliableChannelManager -> MessagingClient (also Waku-owned).
-    ## Hidden so callers can't substitute their own and bypass the
-    ## manager's pipeline.
+    ## Borrowed from the owning `Waku`.
+  sendHandler: SendHandler
+    ## Default egress dispatch for channels created through this manager.
+    ## Constructed at mount time as a closure over `MessagingClient.send`
+    ## so the channel layer itself stays callable-only.
   brokerCtx: BrokerContext
 
 proc new*(
     T: type ReliableChannelManager,
     messagingClient: MessagingClient,
+    sendHandler: SendHandler,
     brokerCtx: BrokerContext = globalBrokerContext(),
 ): Result[T, string] =
   if messagingClient.isNil():
     return err("messaging client is required")
+  if sendHandler.isNil():
+    return err("sendHandler is required")
   ok(
     T(
       channels: initTable[ChannelId, ReliableChannel](),
       messagingClient: messagingClient,
+      sendHandler: sendHandler,
       brokerCtx: brokerCtx,
     )
   )
@@ -61,15 +66,19 @@ proc createReliableChannel*(
     channelId: ChannelId,
     contentTopic: ContentTopic,
     senderId: SdsParticipantID,
+    sendHandler: SendHandler = nil,
 ): Result[ChannelId, string] =
-  ## Spec entry point. The `MessagingClient` and `rng` the channel needs
-  ## are sourced from the owning `ReliableChannelManager` rather than
-  ## passed per call. Encryption is wired up through the `Encrypt`/
-  ## `Decrypt` request brokers — the application installs its own
-  ## providers (or `setNoopEncryption()`) before traffic flows.
+  ## Spec entry point. The `sendHandler` and `rng` the channel needs are
+  ## sourced from the owning `ReliableChannelManager` rather than passed
+  ## per call. Encryption is wired up through the `Encrypt`/`Decrypt`
+  ## request brokers — the application installs its own providers
+  ## (or `setNoopEncryption()`) before traffic flows.
   ##
   ## Segmentation, SDS and rate-limit configs will eventually be read
   ## from the node's `NodeConfig`. Defaults for now.
+  ##
+  ## `sendHandler` defaults to the manager's default (constructed at mount
+  ## from `MessagingClient.send`); tests pass a fake to bypass the network.
   if self.channels.hasKey(channelId):
     return err("channel already exists: " & channelId)
 
@@ -88,8 +97,14 @@ proc createReliableChannel*(
     epochPeriodSec: DefaultEpochPeriodSec, messagesPerEpoch: DefaultMessagesPerEpoch
   )
 
+  let effectiveSendHandler =
+    if sendHandler.isNil():
+      self.sendHandler
+    else:
+      sendHandler
+
   let chn = ReliableChannel.new(
-    messagingClient = self.messagingClient,
+    sendHandler = effectiveSendHandler,
     channelId = channelId,
     contentTopic = contentTopic,
     senderId = senderId,
diff --git a/tests/channels/test_reliable_channel_send_receive.nim b/tests/channels/test_reliable_channel_send_receive.nim
index 6e189d1be..85f327859 100644
--- a/tests/channels/test_reliable_channel_send_receive.nim
+++ b/tests/channels/test_reliable_channel_send_receive.nim
@@ -97,7 +97,7 @@ suite "Reliable Channel - ingress":
     if arrived:
       check received.read() == appPayload
 
-    discard await waku.stop()
+    (await waku.stop()).expect("stop")
 
   asyncTest "manager drops unmarked WakuMessage":
     ## Mirror of the above: same content topic, but `meta` is empty
@@ -150,4 +150,176 @@ suite "Reliable Channel - ingress":
     await sleepAsync(100.milliseconds)
     check not fired
 
-    discard await waku.stop()
+    (await waku.stop()).expect("stop")
+
+suite "Reliable Channel - send state machine":
+  asyncTest "MessageSentEvent finalises the channelReqId as Sent":
+    ## Drives the real send pipeline (`send` -> segmentation -> SDS ->
+    ## rate_limit -> encrypt -> dispatch) via a fake `SendHandler` that
+    ## returns a canned `RequestId` instead of hitting the network.
+    ## Emitting the delivery-layer `MessageSentEvent` must drive the
+    ## channel-level state machine through `Confirmed` and produce a
+    ## `ChannelMessageSentEvent` (channel-level terminal event) for the
+    ## `channelReqId` returned by `send()`.
+    const
+      channelId = ChannelId("sm-success-channel")
+      contentTopic = ContentTopic("/reliable-channel/test/sm-success")
+      fakeMsgReqId = RequestId("fake-msg-req-1")
+
+    var waku: Waku
+    var manager: ReliableChannelManager
+    var brokerCtx: BrokerContext
+    lockNewGlobalBrokerContext:
+      brokerCtx = globalBrokerContext()
+      waku = (await createNode(createApiNodeConf())).expect("createNode")
+      waku.mountMessagingClient().expect("mountMessagingClient")
+      waku.mountReliableChannelManager().expect("mountReliableChannelManager")
+      manager = waku.reliableChannelManager
+
+    setNoopEncryption()
+
+    var sendCalls = 0
+    let fakeSend: SendHandler = proc(
+        env: MessageEnvelope
+    ): Future[Result[RequestId, string]] {.async: (raises: [CatchableError]), gcsafe.} =
+      sendCalls.inc
+      return ok(fakeMsgReqId)
+
+    discard manager
+      .createReliableChannel(
+        channelId, contentTopic, SdsParticipantID("local"), sendHandler = fakeSend
+      )
+      .expect("createReliableChannel")
+
+    let sentFut = newFuture[RequestId]("channel-sent")
+    discard ChannelMessageSentEvent
+      .listen(
+        brokerCtx,
+        proc(evt: ChannelMessageSentEvent) {.async: (raises: []).} =
+          if not sentFut.finished() and evt.channelId == channelId:
+            sentFut.complete(evt.requestId)
+        ,
+      )
+      .expect("listen ChannelMessageSentEvent")
+
+    let channelReqId = manager.send(channelId, "hello".toBytes()).expect("send")
+
+    let dispatchDeadline = Moment.now() + 1.seconds
+    while Moment.now() < dispatchDeadline and sendCalls == 0:
+      await sleepAsync(5.milliseconds)
+    check sendCalls == 1
+
+    waku_message_events.MessageSentEvent.emit(
+      brokerCtx,
+      waku_message_events.MessageSentEvent(requestId: fakeMsgReqId, messageHash: ""),
+    )
+
+    let finalised = await sentFut.withTimeout(1.seconds)
+    check finalised
+    if finalised:
+      check sentFut.read() == channelReqId
+
+    (await waku.stop()).expect("stop")
+
+  asyncTest "two independent channelReqIds are finalised independently":
+    ## Two `send()` calls -> two independent `channelReqId`s, each with
+    ## one segment under the current segmentation skeleton
+    ## (`performSegmentation` always emits exactly one segment). The
+    ## fake `SendHandler` returns distinct `messagingReqId`s; finalising
+    ## the first emits `ChannelMessageSentEvent` for its `channelReqId`,
+    ## finalising the second as a failure emits `ChannelMessageErrorEvent`
+    ## for the other.
+    const
+      channelId = ChannelId("sm-multi-channel")
+      contentTopic = ContentTopic("/reliable-channel/test/sm-multi")
+
+    var waku: Waku
+    var manager: ReliableChannelManager
+    var brokerCtx: BrokerContext
+    lockNewGlobalBrokerContext:
+      brokerCtx = globalBrokerContext()
+      waku = (await createNode(createApiNodeConf())).expect("createNode")
+      waku.mountMessagingClient().expect("mountMessagingClient")
+      waku.mountReliableChannelManager().expect("mountReliableChannelManager")
+      manager = waku.reliableChannelManager
+
+    setNoopEncryption()
+
+    var msgReqIds: seq[RequestId]
+    let fakeSend: SendHandler = proc(
+        env: MessageEnvelope
+    ): Future[Result[RequestId, string]] {.async: (raises: [CatchableError]), gcsafe.} =
+      let id = RequestId("fake-msg-req-" & $(msgReqIds.len + 1))
+      msgReqIds.add(id)
+      return ok(id)
+
+    discard manager
+      .createReliableChannel(
+        channelId, contentTopic, SdsParticipantID("local"), sendHandler = fakeSend
+      )
+      .expect("createReliableChannel")
+
+    let sentFut = newFuture[RequestId]("channel-sent")
+    let erroredFut = newFuture[RequestId]("channel-errored")
+    discard ChannelMessageSentEvent
+      .listen(
+        brokerCtx,
+        proc(evt: ChannelMessageSentEvent) {.async: (raises: []).} =
+          if not sentFut.finished() and evt.channelId == channelId:
+            sentFut.complete(evt.requestId)
+        ,
+      )
+      .expect("listen ChannelMessageSentEvent")
+    discard ChannelMessageErrorEvent
+      .listen(
+        brokerCtx,
+        proc(evt: ChannelMessageErrorEvent) {.async: (raises: []).} =
+          if not erroredFut.finished() and evt.channelId == channelId:
+            erroredFut.complete(evt.requestId)
+        ,
+      )
+      .expect("listen ChannelMessageErrorEvent")
+
+    let channelReqId1 = manager.send(channelId, "first".toBytes()).expect("send 1")
+    let channelReqId2 = manager.send(channelId, "second".toBytes()).expect("send 2")
+
+    let dispatchDeadline = Moment.now() + 1.seconds
+    while Moment.now() < dispatchDeadline and msgReqIds.len < 2:
+      await sleepAsync(5.milliseconds)
+    check msgReqIds.len == 2
+
+    waku_message_events.MessageSentEvent.emit(
+      brokerCtx,
+      waku_message_events.MessageSentEvent(requestId: msgReqIds[0], messageHash: ""),
+    )
+    let sentArrived = await sentFut.withTimeout(1.seconds)
+    check sentArrived
+    if sentArrived:
+      check sentFut.read() == channelReqId1
+    ## The second `channelReqId` must NOT have finalised yet — its
+    ## segment is still `InFlight`.
+    check not erroredFut.finished()
+
+    waku_message_events.MessageErrorEvent.emit(
+      brokerCtx,
+      waku_message_events.MessageErrorEvent(
+        requestId: msgReqIds[1], messageHash: "", error: "synthetic"
+      ),
+    )
+    let erroredArrived = await erroredFut.withTimeout(1.seconds)
+    check erroredArrived
+    if erroredArrived:
+      check erroredFut.read() == channelReqId2
+
+    (await waku.stop()).expect("stop")
+
+  asyncTest "TODO: channelReqId not pruned until ALL its segments are final":
+    ## Placeholder for the multi-sibling prune rule. Today's
+    ## `performSegmentation` (segmentation skeleton) always emits
+    ## exactly one segment per `send()`, so multiple siblings under one
+    ## `channelReqId` cannot be produced through the real pipeline.
+    ## Implement once segmentation does real chunking: send a payload
+    ## larger than `DefaultSegmentSizeBytes`, capture the N
+    ## `messagingReqId`s from a fake `SendHandler`, finalise some, and
+    ## assert prune only fires once every sibling is final.
+    skip()
diff --git a/tests/waku_relay/test_wakunode_relay.nim b/tests/waku_relay/test_wakunode_relay.nim
index da79e2a09..a687119bd 100644
--- a/tests/waku_relay/test_wakunode_relay.nim
+++ b/tests/waku_relay/test_wakunode_relay.nim
@@ -698,13 +698,6 @@ suite "WakuNode - Relay":
     node.unsubscribe((kind: ContentUnsub, topic: contentTopicB)).isOkOr:
       assert false, "Failed to unsubscribe to topic: " & $error
 
-    check node.wakuRelay.isSubscribed(shard)
-
-    node.unsubscribe((kind: ContentUnsub, topic: contentTopicA)).isOkOr:
-      assert false, "Failed to unsubscribe to topic: " & $error
-    node.unsubscribe((kind: ContentUnsub, topic: contentTopicC)).isOkOr:
-      assert false, "Failed to unsubscribe to topic: " & $error
-
     ## After unsubcription, the node should not be subscribed to the shard anymore
     check not node.wakuRelay.isSubscribed(shard)
 
diff --git a/waku/api/api.nim b/waku/api/api.nim
index e0e7d4d7b..24049002b 100644
--- a/waku/api/api.nim
+++ b/waku/api/api.nim
@@ -50,27 +50,4 @@ proc send*(
     w: Waku, envelope: MessageEnvelope
 ): Future[Result[RequestId, string]] {.async.} =
   ?checkApiAvailability(w)
-
-  let isSubbed =
-    w.node.subscriptionManager.isSubscribed(envelope.contentTopic).valueOr(false)
-  if not isSubbed:
-    info "Auto-subscribing to topic on send", contentTopic = envelope.contentTopic
-    w.node.subscriptionManager.subscribe(envelope.contentTopic).isOkOr:
-      warn "Failed to auto-subscribe", error = error
-      return err("Failed to auto-subscribe before sending: " & error)
-
-  let requestId = RequestId.new(w.rng)
-
-  let deliveryTask = DeliveryTask.new(requestId, envelope, w.brokerCtx).valueOr:
-    return err("API send: Failed to create delivery task: " & error)
-
-  info "API send: scheduling delivery task",
-    requestId = $requestId,
-    pubsubTopic = deliveryTask.pubsubTopic,
-    contentTopic = deliveryTask.msg.contentTopic,
-    msgHash = deliveryTask.msgHash.to0xHex(),
-    myPeerId = w.node.peerId()
-
-  asyncSpawn w.messagingClient.sendService.send(deliveryTask)
-
-  return ok(requestId)
+  return await w.messagingClient.send(envelope)
diff --git a/waku/api/types.nim b/waku/api/types.nim
index 9eae503c8..2b7edd616 100644
--- a/waku/api/types.nim
+++ b/waku/api/types.nim
@@ -11,6 +11,10 @@ type
     contentTopic*: ContentTopic
     payload*: seq[byte]
     ephemeral*: bool
+    meta*: seq[byte]
+      ## Opaque wire-format marker carried on the underlying WakuMessage.
+      ## Higher layers (e.g. Reliable Channel) stamp this so peers can route
+      ## ingress traffic to their corresponding layer. Empty by default.
 
   RequestId* = distinct string
 
@@ -34,12 +38,18 @@ proc init*(
     contentTopic: ContentTopic,
     payload: seq[byte] | string,
     ephemeral: bool = false,
+    meta: seq[byte] = @[],
 ): MessageEnvelope =
   when payload is seq[byte]:
-    MessageEnvelope(contentTopic: contentTopic, payload: payload, ephemeral: ephemeral)
+    MessageEnvelope(
+      contentTopic: contentTopic, payload: payload, ephemeral: ephemeral, meta: meta
+    )
   else:
     MessageEnvelope(
-      contentTopic: contentTopic, payload: payload.toBytes(), ephemeral: ephemeral
+      contentTopic: contentTopic,
+      payload: payload.toBytes(),
+      ephemeral: ephemeral,
+      meta: meta,
     )
 
 proc toWakuMessage*(envelope: MessageEnvelope): WakuMessage =
@@ -48,6 +58,7 @@ proc toWakuMessage*(envelope: MessageEnvelope): WakuMessage =
     contentTopic: envelope.contentTopic,
     payload: envelope.payload,
     ephemeral: envelope.ephemeral,
+    meta: envelope.meta,
     timestamp: getNowInNanosecondTime(),
   )
 
diff --git a/waku/factory/waku.nim b/waku/factory/waku.nim
index d4aef252b..ee70cf713 100644
--- a/waku/factory/waku.nim
+++ b/waku/factory/waku.nim
@@ -377,8 +377,15 @@ proc mountReliableChannelManager*(waku: Waku): Result[void, string] =
     return err("reliable channel manager requires a mounted messaging client")
   if waku.node.started:
     return err("cannot mount reliable channel manager on a started node")
+
+  let messagingClient = waku.messagingClient
+  let defaultSendHandler: SendHandler = proc(
+      envelope: MessageEnvelope
+  ): Future[Result[RequestId, string]] {.async: (raises: [CatchableError]), gcsafe.} =
+    return await messagingClient.send(envelope)
+
   waku.reliableChannelManager = ReliableChannelManager.new(
-    waku.messagingClient, waku.brokerCtx
+    messagingClient, defaultSendHandler, waku.brokerCtx
   ).valueOr:
     return err("could not create reliable channel manager: " & $error)
   return ok()
diff --git a/waku/messaging_client.nim b/waku/messaging_client.nim
index 32cb64a9e..1fc4deb3c 100644
--- a/waku/messaging_client.nim
+++ b/waku/messaging_client.nim
@@ -1,7 +1,17 @@
 import results, chronos
-import ./node/waku_node, ./node/delivery_service/[recv_service, send_service]
+import chronicles
+import
+  ./api/types,
+  ./node/[
+    waku_node,
+    subscription_manager,
+    delivery_service/recv_service,
+    delivery_service/send_service,
+    delivery_service/send_service/delivery_task,
+  ]
 
 type MessagingClient* = ref object
+  node: WakuNode
   sendService*: SendService
   recvService*: RecvService
   started: bool
@@ -11,7 +21,7 @@ proc new*(
 ): Result[T, string] =
   let sendService = ?SendService.new(useP2PReliability, node)
   let recvService = RecvService.new(node)
-  ok(T(sendService: sendService, recvService: recvService))
+  ok(T(node: node, sendService: sendService, recvService: recvService))
 
 proc start*(self: MessagingClient): Result[void, string] =
   if self.started:
@@ -27,3 +37,27 @@ proc stop*(self: MessagingClient) {.async.} =
   await self.sendService.stopSendService()
   await self.recvService.stopRecvService()
   self.started = false
+
+proc send*(
+    self: MessagingClient, envelope: MessageEnvelope
+): Future[Result[RequestId, string]] {.async.} =
+  ## High-level messaging API send. Auto-subscribes to the content topic
+  ## (so the local node sees its own gossipsub broadcast), builds a
+  ## `DeliveryTask`, and hands it to the send service. Returns the request
+  ## id the caller can correlate with `MessageSentEvent` / `MessageErrorEvent`.
+  let isSubbed =
+    self.node.subscriptionManager.isSubscribed(envelope.contentTopic).valueOr(false)
+  if not isSubbed:
+    info "Auto-subscribing to topic on send", contentTopic = envelope.contentTopic
+    self.node.subscriptionManager.subscribe(envelope.contentTopic).isOkOr:
+      warn "Failed to auto-subscribe", error = error
+      return err("Failed to auto-subscribe before sending: " & error)
+
+  let requestId = RequestId.new(self.node.rng)
+
+  let deliveryTask = DeliveryTask.new(requestId, envelope, self.node.brokerCtx).valueOr:
+    return err("MessagingClient.send: Failed to create delivery task: " & error)
+
+  asyncSpawn self.sendService.send(deliveryTask)
+
+  return ok(requestId)
diff --git a/waku/node/delivery_service/send_service/send_service.nim b/waku/node/delivery_service/send_service/send_service.nim
index 242dfc111..e60b26124 100644
--- a/waku/node/delivery_service/send_service/send_service.nim
+++ b/waku/node/delivery_service/send_service/send_service.nim
@@ -26,7 +26,7 @@ logScope:
 # This useful util is missing from sequtils, this extends applyIt with predicate...
 template applyItIf*(varSeq, pred, op: untyped) =
   for i in low(varSeq) .. high(varSeq):
-    let it {.inject.} = varSeq[i]
+    var it {.inject.} = varSeq[i]
     if pred:
       op
       varSeq[i] = it