Docker Desktop 4.95.0 / Windows WSL 2: persistent engine-starting state after initial daemon readiness and proxy dial deadline #756

Bug report

Environment

• Docker Desktop 4.95.0.242831 on Windows.
• Docker CLI 29.8.2.
• WSL 2.7.13.0; Docker's WSL distribution is reported as Running, version 2.
• Related Windows services are reported as Running. These observations do not establish Docker API readiness.

Observed behavior

Existing startup logs show containerd boot at 06:22:25.127 UTC, followed by dockerd startup and a successful daemon ping at approximately 06:22:25.970 UTC. A later forwards/list request encounters a deadline. At 06:29:56 UTC, apiproxy reports HTTP 500 for a metadata request concerning an existing container (last known stopped; its current state is unverified):

dialing [private VM endpoint]:2376: context deadline exceeded

The Docker Desktop interface later remains at “Starting the Docker Engine…”. A subsequent CLI container-list request produced no output during a bounded observation before its client was cancelled. Current container and workload states could not be confirmed.

Previous diagnostic actions and limits

• One docker desktop restart --timeout 60 invocation had uncertain completion; its waiting client was cancelled and the restart was not repeated.
• One local diagnostic gather, without upload, was cancelled after approximately 4 minutes 56 seconds by our observation limit. The helper advertises a limit of up to 15 minutes. No ZIP was produced before cancellation; this is not evidence that gathering cannot finish.
• Read-only inspection of existing raw log lines exposed the internal error above.
• No reset, prune, forced process termination, firewall or routing changes, manual internal endpoint probing, or container/volume deletion was performed during this investigation.

Expected behavior and request

Docker should either reach a usable engine-ready state or provide an actionable startup error. The original cause and a reliable reproduction sequence remain unknown. We are not claiming container corruption or a confirmed Docker product defect.

What supported, non-mutating diagnostic observation would distinguish engine unavailability, a proxy/forwarder problem, or an already-expired shared context? Please specify the collection command or documented interface, expected output, duration and side effects. The Docker instance is shared with other workloads, so we need to assess impact and preserve persistent data before any further maintenance.

This report contains a sanitized summary only. No diagnostic bundle, raw logs, container identifiers, local paths, source datasets, credentials or account details are attached. No further collection or upload has been initiated.

Platform

Windows

Version information

Docker Desktop: 4.95.0.242831
Docker CLI: 29.8.2
Windows
WSL: 2.7.13.0
Docker WSL distribution: version 2

Diagnostics ID

No response