fix(preprod): Use consistent external_id for sibling artifacts to prevent duplicate status checks (EME-610)

[runningcode]

Summary

Fixes EME-610: Two related issues causing duplicate status checks and duplicate rows in GitHub check runs for preprod artifacts.

This PR addresses both problems described by Trevor in EME-610:

1. Multiple separate "Size Analysis" check runs appearing on GitHub
2. Duplicate rows appearing within a status check after reprocessing/retriggering CI

Problem 1: Duplicate Check Runs on GitHub

Issue

When multiple sibling artifacts (e.g., iOS, Android, Web in a monorepo) completed processing for the same commit, each artifact would create its own separate GitHub check run because they used different external_id values.

Result: 3 separate "Size Analysis" checks on GitHub instead of 1 unified check.

Root Cause

Each artifact was using its own ID as the external_id:

external_id=str(preprod_artifact.id)  # Each artifact has different ID

GitHub identifies check runs by the tuple: (repo, sha, check_name, external_id)

Different external_id values → GitHub treats them as separate checks

Solution

Use the first sibling's ID consistently as external_id across all sibling artifacts:

first_artifact_id = str(all_artifacts[0].id)  # All siblings use same ID

Now all sibling artifacts share the same external_id, so GitHub recognizes them as updates to the same check run.

Files Changed:

• src/sentry/preprod/vcs/status_checks/size/tasks.py - Use consistent external_id
• tests/.../test_status_checks_tasks.py - Added test test_create_preprod_status_check_task_external_id_consistency

---

Problem 2: Duplicate Rows After Reprocessing

Issue

When CI is retriggered or artifacts are reprocessed, new artifacts are created with the same app_id as existing artifacts. The status check would then show ALL artifacts, resulting in duplicate rows.

Trevor's quote: "The status check itself isn't duplicated...The issue is the duplicated content inside the status check, after re-running it went from 2x builds -> 4x builds."

Example:

• Initial upload: iOS-old, Android-old → 2 rows in status check ✓
• Reprocess: iOS-new, Android-new created (4 total artifacts for same commit)
• Status check now shows: iOS-old, iOS-new, Android-old, Android-new → 4 rows ❌

Root Cause

get_sibling_artifacts_for_commit() returned ALL artifacts for a commit without deduplication:

return PreprodArtifact.objects.filter(
    commit_comparison=self.commit_comparison,
    project__organization_id=self.project.organization_id,
).order_by("app_id")

Solution

Deduplicate siblings by app_id, matching the behavior in the Emerge codebase (BaseBuildUtils.ts:findSiblingBuilds):

1. Group all artifacts by app_id
2. For each app_id group:
   • If the calling artifact is in that group → use it
   • Otherwise → use the earliest (oldest) artifact for that app_id
3. Result: Max 1 artifact per app_id

Example with fix:

• When iOS-new triggers: Returns [iOS-new, Android-old] → 2 rows ✓
• When Android-new triggers: Returns [iOS-old, Android-new] → 2 rows ✓

Files Changed:

• src/sentry/preprod/models.py - Added deduplication logic to get_sibling_artifacts_for_commit()
• tests/.../test_status_checks_tasks.py - Added test test_sibling_deduplication_after_reprocessing

---

Additional Changes

• src/sentry/integrations/github/client.py: Added update_check_run() method for future flexibility (not currently used but documents GitHub API capability)

---

Testing

Added comprehensive tests:

1. test_create_preprod_status_check_task_external_id_consistency: Verifies all sibling artifacts use the same external_id
2. test_sibling_deduplication_after_reprocessing: Verifies deduplication prevents showing all 4 artifacts after reprocessing

All 18 existing tests still pass ✅

---

Verification

To verify the fix works:

1. Run bin/preprod/reproduce_duplicate_checks_bug multiple times
2. Check the GitHub commit - should see 1 "Size Analysis" check (not 3)
3. The check should show 2 rows consistently (not 4 after retriggering)

[linear]

EME-610 Reprocessing status check didn't work

[codecov]

Codecov Report

❌ Patch coverage is 84.00000% with 4 lines in your changes missing coverage. Please review.
✅ All tests successful. No failed tests found.

Files with missing lines	Patch %	Lines
src/sentry/integrations/github/client.py	33.33%	2 Missing ⚠️
src/sentry/preprod/vcs/status_checks/size/tasks.py	81.81%	2 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff             @@
##           master   #103521      +/-   ##
===========================================
+ Coverage   78.03%    80.60%   +2.56%     
===========================================
  Files        9245      9250       +5     
  Lines      394973    395098     +125     
  Branches    25190     25190              
===========================================
+ Hits       308215    318461   +10246     
+ Misses      86327     76206   -10121     
  Partials      431       431