Orchestrata Ultimate › Plans › briefs › kimi-c1
Kimi task C1 — failure census of the 2026-07-22 full-suite run
Source: plans/briefs/kimi-c1.md at 8a2cb42ad · last changed 2026-07-27 · first committed 2026-07-22 · 25 lines.
Prose pending.
Plan text
Kimi task C1 — failure census of the 2026-07-22 full-suite run
Read plans/parallel-collision-fix-plan.md (Workstream C1) for context. This is a
classification task: no code changes, no fixes.
Input: plans/briefs/full_suite_run2.json — output of scripts/run_tests.py --json
(27,187 passed / 41 failed / 15 errors, 19 failed modules). Each module entry under
"modules" carries returncode, stdout, stderr.
Task: for every module with returncode != 0, read its captured output and classify EVERY
individual test failure/error into exactly one of:
port-race— bind failures on KG_GRPC_PORT 50052 (Failed to bind to address), connection refused to a KG gRPC server that lost the bind race, or cascades directly caused by these.file-dir-constraint— Neo4jConstraintValidationFailedonFile.path/Directory.path(MERGE-key-vs-constraint-key mismatch), or cascades directly caused by these.other— anything that fits neither. Quote the decisive error line verbatim.
Deliverable: write plans/briefs/kimi-c1-result.md containing:
- A table: module | test id | class | decisive error line (truncated to 120 chars).
- Totals per class; they must sum to 41 failures + 15 errors (state the sum explicitly).
- A final section
## OTHER-class findings— everyotherentry with full error text. If this section is non-empty, start it with the lineESCALATE: third failure class present.
Do not modify any other file. Do not run the test suite.