[eirsep]

Description

Follow-up to #2315, addressing @lezzago's review. That PR recorded the wrong cause and carried a
second change that was not needed.

The explanation was wrong. PER_EXECUTION actions do advance lastNotificationTime — that branch
adds every deduped, new and completed alert to alertsToUpdate
(BucketLevelMonitorRunner.kt:395-409). An all-PER_EXECUTION trigger was never the problem, and the
comment #2315 left in the test said it was.

The actual cause is that randomActionsForBucketLevelTrigger(min = 1) returned an empty list, and
randomBucketLevelTrigger replaces an empty actions list with freshly generated ones
(TestHelpers.kt:370):

actions = if (actions.isEmpty()) randomActionsForBucketLevelTrigger(destinationId = destinationId) else actions

Those regenerated actions never pass through the rewrite the test applies, so a PER_ALERT policy
without COMPLETED reaches the trigger. At the seed from the original failure they were
[PerAlertActionScope{DEDUPED}, PerAlertActionScope{}].

Change

Making min a floor — the other half of #2315 — fixes this on its own, so the PER_ALERT action that PR
prepended is removed. Besides being unnecessary, it guaranteed a PER_ALERT action carrying COMPLETED
on every run, which stopped this test from ever exercising an all-PER_EXECUTION trigger: it narrowed
coverage to buy nothing. It also duplicated the ActionExecutionPolicy expression within one statement.

The test comment now states the real cause, and the helper's KDoc warns that an empty result is not
harmless for callers that post-process it, since that is the trap behind the original failure.

Testing

With the prepend removed:

	result
seed from the original failing CI job (F63F1F740575292F, locale ki, timezone ACT)	pass
10 further random seeds	10/10 pass

ktlint clean. No production code is touched.

Still open

maxOf(min, 1) means an explicit min = 0 can no longer yield zero actions, which the PR bot flagged
on #2315. Given TestHelpers.kt:370 treats an empty list as "regenerate", zero actions is not
reachable through randomBucketLevelTrigger either way. Making it genuinely reachable means stopping
that silent regeneration, which changes every caller's action distribution and belongs in its own
change rather than here.

Check List

• New functionality includes testing.
• New functionality has been documented.
• API changes companion pull request created.
• Commits are signed per the DCO using --signoff.
• Public documentation issue/PR created.

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.
For more information on following Developer Certificate of Origin and signing off your commits, please check here.

[github-actions]

PR Reviewer Guide 🔍

Here are some key observations to aid the review process:

🧪 PR contains tests

🔒 No security concerns identified

✅ No TODO sections

🔀 No multiple PR themes

⚡ No major issues detected

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 67.71%. Comparing base (778e7ce) to head (3855ec0).
⚠️ Report is 254 commits behind head on main.

❌ Your project check has failed because the head coverage (67.71%) is below the target coverage (70.00%). You can increase the head coverage or adjust the target coverage.

Additional details and impacted files

@@             Coverage Diff              @@
##               main    #2316      +/-   ##
============================================
- Coverage     67.72%   67.71%   -0.02%     
- Complexity      105     1795    +1690     
============================================
  Files           160      169       +9     
  Lines         10343    14269    +3926     
  Branches       1522     2077     +555     
============================================
+ Hits           7005     9662    +2657     
- Misses         2672     3728    +1056     
- Partials        666      879     +213     

☔ 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.

[lezzago]

nit: Regenerated actions can also lack DEDUPED (e.g. PerAlertActionScope{NEW}), and then the test fails earlier, at "Active alert was not updated", rather than at the completed-alert check. Maybe: "...so a PER_ALERT policy missing DEDUPED or COMPLETED can reach the trigger and leave lastNotificationTime untouched."

[lezzago]

nit: The non-empty guarantee the comment above depends on lives in the helper's maxOf(min, 1), which the description suggests revisiting. An assertTrue(actions.isNotEmpty()) after building actions would make this test fail clearly instead of flaking again if that helper changes.