fix(analytics): surface polling-loop errors so schema drift can't loop silently (SDK-78)

[tylerjroach]

Summary

start_polling_for_definitions! caught StandardError and dispatched only to @error_handler, whose default Mixpanel::ErrorHandler#handle is a no-op. Schema drift (NoMethodError, JSON::ParserError, TypeError, …) would loop forever undetected unless the user had configured a custom handler — exactly the audit's concern.

Add a log_polling_error helper that always warns to $stderr before dispatching to @error_handler. Visibility no longer depends on configuration; operators see the failure in their logs the moment the next poll hits a bad payload.

Why a warn and not a re-raise

Matches the de facto convention across every other Mixpanel SDK polling loop:

SDK	Behavior
mixpanel-python	logger.exception("Failed to fetch feature flag definitions"), continue
mixpanel-java	logger.log(Level.WARNING, "Failed to fetch flag definitions", e), continue
mixpanel-go	log.Printf("Error polling for flag definitions: %v", err), continue
mixpanel-node	this.logger?.error(...), continue

Crashing the polling thread on the first bad payload would freeze local evaluation at whatever variants were last cached — strictly worse than continuing. The audit's complaint was about detection, not retry policy.

Context

Linear: SDK-78. Source: Feature Flags & OpenFeature SDK Audit Findings. Ruby was the only affected SDK — the rest already log unconditionally.

Independent of SDK-81 (the polling-thread-safety fix); changes are in different lines of the same begin/rescue block and will merge cleanly in either order.

Test plan

• Full flags spec suite (46 examples) passes
• New "warns to stderr when fetch raises and no error_handler is configured" exercises the audit-failure mode directly (provider built with nil error_handler + 500 response, asserts the warn lands on stderr)
• New "surfaces unexpected errors (schema drift) instead of swallowing them silently" injects a NoMethodError via allow(...).to receive(:fetch_flag_definitions), asserts both the error_handler dispatch AND the stderr warning fire

🤖 Generated with Claude Code

[linear-code]

SDK-78

[codecov]

Codecov Report

❌ Patch coverage is 80.00000% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 96.66%. Comparing base (e5edf71) to head (54a7b06).

Files with missing lines	Patch %	Lines
lib/mixpanel-ruby/flags/local_flags_provider.rb	80.00%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #156      +/-   ##
==========================================
+ Coverage   96.64%   96.66%   +0.01%     
==========================================
  Files          14       14              
  Lines         656      659       +3     
==========================================
+ Hits          634      637       +3     
  Misses         22       22              

Flag	Coverage Δ
openfeature	100.00% <ø> (ø)

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.

[tylerjroach]

Folded into #155 — both fixes touch the same begin/rescue block in start_polling_for_definitions!, simpler as one PR.