[Incident][108dev] Intermittent 500 on essays endpoints from SWA frontend

[Cataldir]

Summary

A new frontend incident was observed during live demo prep: essays screen failed with 500 responses on APIM routes.

Observed from browser console (SWA):

• GET /api/essays/resources -> 500
• GET /api/essays/essays -> 500
• GET /api/essays/assemblies -> 500

Scope

• Environment: 108dev
• Source: SWA frontend (
  ice-flower-0be65ed0f.1.azurestaticapps.net)
• Impact: Essays page displays no available cases and cannot load data.

Verification done now

Re-check from command line against same APIM routes:

• Preflight CORS for all three routes -> 200 with Access-Control-Allow-Origin set to SWA origin.
• Functional GET for all three routes -> 200 with payload.

This indicates an intermittent live regression window (transient backend/runtime drift or in-flight deployment state), not a persistent outage at check time.

Actions tracked

• Live remediation actions and drift context are tracked in [Incident][108dev] APIM/SWA integration regressions: CORS + configuration themes 500 #139.
• This issue is opened to track intermittent essays endpoint instability specifically.

Next steps

• Add periodic synthetic checks for /api/essays/resources|essays|assemblies from SWA origin.
• Backport runtime drift protections to IaC/workflow to prevent env resets.
• Correlate failures with active workflow/deploy windows and revision rollouts.

Related: #139

[Cataldir]

Issue #141 live correction update (new regression from frontend screenshot):

Root cause identified:

• Essay essay-econ-001 had assembly_id link, but corresponding assembly documents were missing in essays storage.
• API /api/essays/grader/interaction returned 400 with:
  • Unable to resolve assembly for case_id 'essay-econ-001'.

Live correction applied:

• Recreated missing assembly records for all seeded essays via /api/essays/assemblies with matching IDs:
  • asm-essay-fund1-001
  • asm-essay-fund2-001
  • asm-essay-fund2-002
  • asm-essay-em-001
  • asm-essay-em-002
  • asm-essay-econ-001
  • asm-essay-econ-002
  • asm-essay-econ-003

Verification:

• Re-tested the exact failing flow for essay-econ-001.
• POST /api/essays/grader/interaction now returns 200 (fallback evaluation payload), no unresolved assembly error.

Note:

• Agent provisioning path is still under stabilization, so evaluation currently falls back gracefully when no provisioned agents are present.
• This correction restores frontend flow continuity and removes the blocking 400 error.