[NA] [E2E] fix: increase online scoring test timeouts for GitHub-hosted runners

[AndreiCautisanu]

Details

Online scoring E2E tests were consistently timing out (60s) on GitHub-hosted runners in post-merge CI while passing on self-hosted runners. Investigation via Allure TestOps showed the scoring pipeline (rule activation → LLM API call → score storage) takes longer on resource-constrained ubuntu-latest runners compared to self-hosted automation-q5pjv-* pods.

• Test timeout increased from 60s to 120s
• Polling attempts increased from 15 to 25
• Page refresh wait increased from 2s to 3s

Change checklist

• User facing
• Documentation update

Issues

• NA — flaky test fix for post-merge CI

AI-WATERMARK

AI-WATERMARK: yes

• Tools: Claude Code
• Model(s): Claude Opus 4.6
• Scope: investigation + fix
• Human verification: code review, will verify in next post-merge run

Testing

Verified by analyzing Allure TestOps data:

• Post-merge [OPIK-374] Add bulk actions to all tables #804 (launch 53489): 6/8 online scoring tests timed out at ~62s
• Post-merge [OPIK-515] implement opik client auth_check method #805 (launch 53491): 2/8 timed out at ~63s
• Local Happy Paths (launch 53456): all passed in 28-51s
• New timeout budget: ~15s setup + 25 retries × ~3.5s ≈ 102s, within 120s limit

Documentation

N/A

[baz-reviewer]

Should we replace PAGE_REFRESH_TIMEOUT = 3000 and any page.waitForTimeout uses with retry-capable assertions or helper waits per .agents/skills/playwright-e2e/test-conventions.md, and remove the sleep in the moderation scoring retry loop?

_{Finding type: AI Coding Guidelines | Severity: 🟢 Low}

---

Want Baz to fix this for you? Activate Fixer

Other fix methods

Prompt for AI Agents:

Before applying, verify this suggestion against the current code. In
tests_end_to_end/typescript-tests/tests/online-scoring/online-scoring.spec.ts around
lines 10 to 10, the PAGE_REFRESH_TIMEOUT constant is set to 3000ms which introduces a
hard-coded delay >2s; per test conventions remove or reduce this constant and do not use
it for page.waitForTimeout. Replace any uses of PAGE_REFRESH_TIMEOUT and any
page.waitForTimeout calls (search around lines 120-140 and specifically 129-133) with
retry-capable helpers such as expect(...).toBeVisible({ timeout: 2000 }) or
helperClient.waitUntil/polling helpers. Also in the moderation scoring retry loop around
lines 106 to 113 (and any loop using sleeps), remove the sleep-based waits and implement
polling using expect/assert-with-timeout or a helper that retries until the moderation
column appears or a total timeout is reached, ensuring no single hard-coded pause
exceeds 2s.