Agent Teams: production coordination patterns (crash recovery, file transactions, persistent backlog)

Preflight Checklist

• [x] I have searched existing requests and this feature hasn't been requested yet
• [x] This is a single feature request (not multiple features)

Problem Statement

When running multiple Claude Code agents on a shared codebase in production, several failure modes emerge: context loss after compaction, file conflicts between agents, no crash recovery, no persistent task backlog between sessions, and no cross-session knowledge retention. These become blockers at scale.

Proposed Solution

Full writeup with 8 pain points, specs, and invariants based on production experience:
https://gist.github.com/sigalovskinick/6cc1cef061f76b7edd198e0ebc863397

Happy to share artifacts (DB schema, event contracts, pipeline templates) or write an ADR-style proposal for any specific area.

Alternative Solutions

Built a dual-orchestrator coordination layer (Claude + Kimi) that solves these — persistent workers, file transactions, 5-phase workflow, autonomous backlog. Details in the writeup above.

Priority

Low - Nice to have

Feature Category

Other

Use Case Example

_No response_

Additional Context

_No response_