Flaky TcpConnectionSpec pull-mode test times out waiting for Tcp.Received

[He-Pin]

Summary

The TcpConnectionSpec test named An outgoing connection must respect pull mode timed out while waiting for Tcp.Received during validatePullRequest.

Reproduction

Run the full PR validation on macOS with JDK 25:

sbt validatePullRequest

The failure was observed after the test suite had already exercised several TCP integration tests in the same run.

Observed failure

An outgoing connection must respect pull mode *** FAILED ***
java.lang.AssertionError: assertion failed: timeout (3 seconds) during expectMsgClass waiting for class org.apache.pekko.io.Tcp$Received
  at org.apache.pekko.io.TcpConnectionSpec$$anon$15.$anonfun$new$35(TcpConnectionSpec.scala:419)

Expected behavior

The test should reliably receive Tcp.Received and pass under normal local PR validation load.

Environment

• macOS
• Oracle JDK 25.0.1
• sbt 1.12.13
• Apache Pekko main-based branch at be50923

Notes

This surfaced while validating PR #3030, but the PR only changes GraphInterpreter stage-finalization bookkeeping. The focused GraphInterpreter, lifecycle, port-state, failure-mode, and actor-shell suites passed 126 of 126 tests before this unrelated TCP timeout. The failure therefore looks like a timing-sensitive test flake rather than a regression from PR #3030.

The full validatePullRequest run was stopped after this failure because an unrelated AGENTS.md worktree change caused the all-build filter to select the complete repository test matrix.

[He-Pin]

Focused rerun passed immediately on the same environment:

sbt "actor-tests / Test / testOnly org.apache.pekko.io.TcpConnectionSpec -- -z \"respect pull mode\""

Result: 1 test run, 1 succeeded, 0 failed; the test completed in 474 ms. This strengthens the timing-sensitive flake diagnosis.