[META] Cowork cluster — 2026-06 active sub-issues (extends closed #54891 / #54847)
Background
The original master tracking ticket #54891 (extending #54847) was auto-closed by the stale-bot on 2026-06-01 due to inactivity on the master thread, even though multiple sub-issues in the cluster remain open and accumulate community evidence.
Opening this lightweight tracking ticket to maintain a single discoverable entry point for the cluster, so anyone landing on #54891 can find the active sub-issues without re-stitching the lineage from scratch.
Active sub-issues (as of 2026-06-08)
| # | Title | Status | Latest activity |
|---|---|---|---|
| #64592 | [BUG] Cowork — VM service not running on Windows 11 (fresh repro + workaround; extends #54891 / #61559) | OPEN | 2026-06-06 (ARM64 / Snapdragon X cross-platform evidence + HCS-level root cause data) |
| #64600 | [BUG] Concurrent writes to .claude.json by parallel workers + MCP truncate the file mid-JSON, triggering cascading agent spawn | OPEN | 2026-06-02 (root-cause confirmation + operator-side workarounds) |
| #64596 | [BUG] Write/Edit tools leave NULL-byte tail padding when shortening files (no O_TRUNC) on Windows + virtiofs | OPEN | duplicate-bot warning, not closed |
| #64597 | [BUG] Write/Edit tools slice CJK characters mid-UTF-8-sequence, leaving partial 0xe3/0xe4 lead bytes at file tail | OPEN | duplicate-bot warning, not closed |
| #64598 | [BUG] Read tool serves cached payload after failed Write, masking on-disk corruption (no cache invalidation on write) | OPEN | duplicate-bot warning, not closed |
Closed predecessors (for lineage)
- #54891 — closed 2026-06-01 (stale-bot, not actually resolved)
- #54847 — the original tool-dispatch-stall thread
- #61559 — earlier Cowork VM thread
Themes across the cluster
- Cowork VM service initialization path on Windows 11 (#64592) — platform-independent (host-side) and architecture-independent (x86_64 + ARM64), guest OS never boots
- File-write atomicity / locking gaps (#64596 / #64597 / #64598 / #64600) — Write/Edit tools lack atomic-write semantics, leaving partial / corrupt state on disk; Read tool cache compounds the issue by masking corruption
- Common second-order effect: silent state corruption that the dispatcher / parent agents do not surface as errors
Purpose of this ticket
- Do not duplicate sub-issue discussion here — keep technical detail in each sub-issue
- Provide a single landing point that survives stale-bot for the cluster
- Update the table above as sub-issues close or new related issues are filed
This issue has 1 comment on GitHub. Read the full discussion on GitHub ↗