[wangyinz]

Closes #795.

Replaces the HTTP probe on Dask's scheduler protocol port with one bounded distributed.Client scheduler RPC. Worker and frontend services now wait for a healthy scheduler in all three Dask Compose topologies, while MongoDB dependency behavior remains unchanged.

Validation:

• contract tests: 11 passed, 3 opt-in Docker lifecycle tests skipped; exact master baseline: 10 failed, 1 passed, 3 skipped
• live Dask 2025.10.0 probe: healthy scheduler returned 0 and closed the client; stopped scheduler returned 1 after the two-second timeout
• docker compose config, Black, py_compile, and git diff --check: passed

Independent agent review: PASS with no code changes.

The local Docker daemon was unavailable, so the opt-in full-container lifecycle tests were not claimed as executed; CI is expected to exercise container integration.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 74.34%. Comparing base (726abc7) to head (a87168e).
⚠️ Report is 1 commits behind head on master.

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #938   +/-   ##
=======================================
  Coverage   74.34%   74.34%           
=======================================
  Files         183      183           
  Lines       36506    36506           
  Branches     2540     2540           
=======================================
+ Hits        27140    27141    +1     
+ Misses       8456     8455    -1     
  Partials      910      910           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[wangyinz]

Review blocker: the three required running-topology tests are currently opt-in and no CI job enables them. This PR will remain Draft until a path-filtered Compose job exercises healthy startup, task submission, scheduler loss, unhealthy transition, and dependency gating on the final head.

[wangyinz]

Resolved: a path-filtered Compose job now enables the live suite. The final head passed all 15 tests, including healthy startup and task submission for all three tracked topologies, scheduler-loss health transitions, and dependency restart gating.