fix: Prevent losing a status message when redirecting Actor run logs

[vdusek]

What

ActorClient.call built its two log redirectors inside a single with statement:

with run_client.get_status_message_watcher(to_logger=logger), run_client.get_streamed_log(to_logger=logger):

with A(), B(): enters A before constructing B, so the status watcher's polling thread was already running - and already logging into apify.<actor> runId:<id> - when get_streamed_log called create_redirect_logger on that same logger. That call reconfigures the logger from scratch (propagate = False, removeHandler, addHandler), so a status message emitted in that window went to a logger that was mid-rebuild and was dropped.

Both redirectors are now constructed before either one starts polling, which is what the async call already did.

Why

This surfaced as a flaky unit test - CI run 32853344220, test_actor_call_redirect_logs_to_default_logger_sync on ubuntu-latest / 3.13. Status: RUNNING, Message: Initial message appeared in captured stderr but not in caplog, i.e. the record reached the logger's own handler while propagate was False. Only the sync test ever flaked, because the async call constructs both redirectors up front and so has no such window.

The nondeterminism is in the client, not in the test: under real usage the same window can silently drop a status message.

Test

The new test pins the damaging interleaving with two events instead of relying on thread scheduling, so it is deterministic in both directions - it fails on master and passes with this change, at the same runtime either way.

Verification: the new test fails without the actor.py change and passes with it; full unit suite (911 passed), lint, type check and docstring check all pass. The original flake does not reproduce naturally on an 8-core machine (0 failures in 40 x tests/unit/test_logging.py under --numprocesses=16 before the change), which is why the regression test forces the interleaving rather than hammering for it.

✍️ Drafted by Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 95.18%. Comparing base (514848f) to head (93a3f45).
⚠️ Report is 1 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1036      +/-   ##
==========================================
- Coverage   95.23%   95.18%   -0.06%     
==========================================
  Files          58       58              
  Lines        5436     5438       +2     
==========================================
- Hits         5177     5176       -1     
- Misses        259      262       +3     

Flag	Coverage Δ
integration	91.70% <100.00%> (-0.37%)	⬇️
unit	87.21% <100.00%> (+<0.01%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ 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.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.