fix(podman): tolerate shutdown transport closes

[elezar]

Summary

• increase the Podman driver graceful stop default from 10s to 45s while keeping the existing OPENSHELL_STOP_TIMEOUT override
• classify h2 protocol errors, broken pipes, cancelled streams, reset connections, and EOF as expected only once sandbox/session shutdown is already underway
• keep unexpected transport failures on the existing warning/failure path and update Podman docs for the new default

Related failing job: https://github.com/NVIDIA/OpenShell/actions/runs/30108191582/job/89533801934

Changes

• add shared transport-close classification in openshell-core
• make gateway supervisor/relay teardown logging shutdown-aware using sandbox deletion/session state
• make the in-sandbox supervisor session stop reconnecting/failing on expected local shutdown transport closes
• update Podman stop timeout docs and config defaults

Testing

• cargo fmt --all
• git diff --check
• cargo test -p openshell-core transport_errors
• cargo test -p openshell-server --lib supervisor_session::tests
• cargo test -p openshell-supervisor-process --lib supervisor_session

Skipped pre-commit per handoff request.

Checklist

• Tests added/updated
• Documentation updated
• DCO sign-off included
• Full pre-commit/CI run

[copy-pr-bot]

Auto-sync is disabled for draft pull requests in this repository. Workflows must be run manually.

Contributors can view more details about this message here.

[github-actions]

🌿 Preview your docs: https://nvidia-preview-pr-2498.docs.buildwithfern.com/openshell

[github-actions]

Label test:e2e applied, but pull-request/2498 does not exist yet. A maintainer needs to comment /ok to test bf23aca4ee31fe35ee1764d4b36d57d1ce626acd to mirror this PR. Once the mirror exists, re-apply the label or re-run Branch E2E Checks from the Actions tab.

[elezar]

/ok-to-test bf23aca

[elezar]

/ok-to-test 50c43f1

[elezar]

/ok-to-test 7eeafa1

[drew]

gator-agent

Maintainer Approval Needed

Gator validation and PR monitoring are complete for this head.

Validation: PR #2498 is project-valid because it is a focused Podman/supervisor shutdown fix from a maintainer-author with clear motivation and scoped docs updates.
Head SHA: 7eeafa195ea5a7bddbbaa1729219da55de9d715a
Review: Independent principal-engineer review found no blocking correctness, security, maintainability, or docs/navigation findings.
Docs: Updated under docs/reference/ for the user-facing Podman stop-timeout default change; no Fern navigation change appears required.
Checks: Required gates are green, including DCO, OpenShell / Branch Checks, OpenShell / Helm Lint, and OpenShell / E2E.
E2E: test:e2e is applied and the E2E gate passed. GPU E2E is not required for this change.

Human maintainer approval is now required.

[drew]

gator-agent

Monitoring Complete

Monitoring is complete because this PR has merged.

Final status: PR #2498 was in gator:merge-ready with maintainer approval present and the required gates green before merge. The monitored head was 7eeafa195ea5a7bddbbaa1729219da55de9d715a.

I removed the active gator:* label because there is nothing left for gator to monitor on this PR.

[juhp]

Just commenting, so now (with 0.0.99):

$ time openshell sandbox create --no-keep -- date

Created sandbox: compelling-velvetbr                                                                                                                                                                                                                                                    

Thu Aug  6 06:19:22 UTC 2026
✓ Deleted sandbox compelling-velvetbr

real	0m48.885s
user	0m0.016s
sys		0m0.021s

Grok helped me track it down to this change.