Problem

SlackThreadBindingActor and SignalRSessionActor duplicate ~80 lines of identical pipeline lifecycle code:

• EnsureInitializedAsync() — creates ActorMaterializer, runs pipeline, materializes input/output, registers OutputStreamTerminated continuation
• ReinitializePipelineAsync(string reason) — completes input queue, disposes session, disposes materializer, re-initializes
• PostStop() — identical three-line teardown
• Private message types: OutputStreamTerminated(int Generation, Exception? Cause), ReinitializePipeline(string Reason)

SignalRSessionActor even acknowledges this with the comment: "Mirrors the pattern used by SlackThreadBindingActor".

Proposed Solution

Create an abstract PipelineHostingActor base class in src/Netclaw.Actors/Channels/ that consolidates the shared lifecycle code. Subclasses override:

• Pipeline — the ISessionPipeline to use
• PipelineOptions — channel-specific SessionPipelineOptions
• OnOutput(SessionOutput) — how to route output to the actor's mailbox

Both actors would extend PipelineHostingActor instead of ReceiveActor.

Risk

High — touches Akka actor lifecycle. Requires careful testing of both Slack and SignalR session flows.

Files

• src/Netclaw.Channels/Slack/SlackThreadBindingActor.cs
• src/Netclaw.Daemon/Gateway/SignalRSessionActor.cs
• New: src/Netclaw.Actors/Channels/PipelineHostingActor.cs

Context

Identified during the code quality audit in #302.