GPU process crash while a preview dev server times out — kills Cowork RPC/VM connection

Status Closed — duplicate
Maintainer reply None cached
Activity 1 comment · opened Jul 26, 2026 · closed Aug 25, 2026
  • App version 1.24012.9, Windows 11, MSIX install
  • Sequence observed in logs:
  • 13:35:11 — preview dev server for a project failed to become ready within timeout (unrelated project, different port than this session's own preview, which stayed healthy throughout)
  • 13:35:21GPU process gone: reason: 'crashed', exitCode: 101457950
  • Same second — Cowork's persistent RPC connection dies (failed to read length: EOF, subscriber disconnected) — a symptom of the Electron app itself going down, not an independent Cowork fault
  • 13:36:29 — Cowork sandbox VM's graceful shutdown fails, force-terminated (HcsShutdownComputeSystem HRESULT error), cascading further disconnects
  • 13:37:40 — app restarts fresh
  • GPU info dump shows both GPUs active: false at capture time, glImplementationParts: "(gl=none,angle=none)" — consistent with the GPU process having crashed and not recovered
  • Workaround in use: avoid preview/design panes on flaky dev servers until fixed

View original on GitHub ↗

This issue has 1 comment on GitHub. Read the full discussion on GitHub ↗