Forumics logo Forumics
Issue 01

Back to issue

All Claude Code known issues

CLAIM C116 Supported 2026-11-13

GitHub issue #93482 in anthropics/claude-code, opened 2026-09-10 by GBalunis, records a report that in a cloud Cowork session linked to a desktop device (Claude Desktop 1.49585.0, Windows 11 25H2, claude-opus-5), when device_commit_files writes back to a file that already exists on the linked device, it returns written with no rejections and the file's modification time advances to the moment of the call, but the content on disk is the payload from the previous commit; committing the identical payload a second time lands it; creating a new file is unaffected and correct every time; the report calls this a silent data-integrity failure, rules out a read cache, OneDrive, KB5124008, and the force flag, and offers a suggested area to look at without asserting a root cause.

Verbatim Excerpts from Cited Records
“When a cloud Cowork session writes back to a file that **already exists** on the linked device, `device_commit_files` returns `{"written":[path],"rejected":[]}`, the file's modification time advances to the moment of the call, and **the content on disk is the payload from the previous commit**.”
— GBalunis · anthropics/claude-code (issue 93482 opening body, Summary section) · 2026-09-10T22:02:35Z ·GitHub Issue ↗
“Creating a **new** file is unaffected and correct every time. Only overwrites are affected.”
— GBalunis · anthropics/claude-code (issue 93482 opening body, Summary section) · 2026-09-10T22:02:35Z ·GitHub Issue ↗
“This is a silent data-integrity failure. The session has no signal that anything went wrong, and the file it just "wrote" looks freshly updated to any human or tool inspecting it, because the timestamp is current while the bytes are stale.”
— GBalunis · anthropics/claude-code (issue 93482 opening body, Summary section) · 2026-09-10T22:02:35Z ·GitHub Issue ↗
“After `device_commit_files` returns `written` with no rejections, the file on disk should contain the committed payload.”
— GBalunis · anthropics/claude-code (issue 93482 opening body, What Should Happen section) · 2026-09-10T22:02:35Z ·GitHub Issue ↗
Counter-record: Issue open as of the fetch; 24 comments were read. Skartrek (2026-09-13) confirms the same behavior on Windows Cowork. GBalunis (2026-09-14) confirms it is still present after the Windows fix and a desktop update to 1.52386.6. PensionedITguy (2026-09-16) independently reproduces on Windows 11 x64, Desktop 1.52386.6, on a local SSD. samtuominen (2026-09-17) confirms from Windows 11 ARM64, with four instances across three projects. lukaszkakol (2026-09-18) corroborates on a cloud session linked to a Mac, so not platform- or model-specific. macdonbruce-hub (2026-09-19) reports no longer reproducing on 2.2553.1, then (2026-09-20) corrects that run as a false negative and re-tests with the same staged path reused, reproducing the bug. GBalunis (2026-09-19) reports it is not fixed on 2.2553.1 but the trigger is narrower than assumed. RadicalZephyr (2026-09-20) confirms on Linux desktop with a minimal reproducer. Vicellia (2026-09-23) independently reproduces and adds that the stale re-send overwrites host-side edits made between the two commits. GBalunis (2026-09-23) confirms that destructive finding on 2.7032.0. bdelima (2026-09-24) files a condensed bug report calling it deterministic and high severity. macdonbruce-hub (2026-09-27) reports both cases clean on 2.9939.2, but bdelima (2026-09-28) and GBalunis (2026-09-29, tiebreaker on 2.9939.4) and FangAlpha (2026-09-29) confirm it is still present. aitkeni54 (2026-09-30) adds that a prior commit is not required. GBalunis (2026-10-11) corrects his own earlier consecutive-reuse trigger claim in favour of the timing model. ElMinouh (2026-10-12) confirms on 2.31226.1. rstribrn (2026-10-13) independently reproduces and isolates delay between the container-side edit and the commit as a variable. GBalunis (2026-10-13) confirms that guard observation by controlled A/B. agentgreen-lang (2026-10-23) adds a data point on 3.1623.0. funschooling (2026-11-12) reports it still reproduces on 5.1231.2, seen 2026-11-11. agentgreen-lang (2026-11-12) reports no reproduction across several dozen writes on desktop app 4.23118.0. No maintainer fix and no retraction of the report appear in the surveyed thread. The claim documents only the original report, not a verified current state. Falsification probe: If GitHub issue #93482 is deleted without archive, if the author edits the opening body so the quoted sentences no longer match verbatim, or if the author posts a retraction of the report, this claim drops to Withdrawn. Audit verification: Verified 100% continuous verbatim substring against official GitHub REST API by scripts/verify_claim_verbatim.py.