refactor(onboard): extract sandbox create-failure reporting and create step into modules by Dongni-Yang · Pull Request #6444 · NVIDIA/NemoClaw

Dongni-Yang · 2026-07-08T05:53:40Z

Summary

Extracts cohesive units of the sandbox create/register orchestration out of the ~4,900-line src/lib/onboard.ts entrypoint into focused modules under src/lib/onboard/, following the injected-deps boundary style established by created-sandbox-finalization.ts (#6332). The primary goal is maintainability and independent unit-test coverage for the create path, with intentional safety/behavior refinements discovered during review: redacted create-output failure echoing, preservation of non-zero create-stream status when readiness fails, direct argv spawning for trusted create paths, fail-closed Docker-GPU create-poll side-effect handling, and redacted trace reporting for poll/readiness errors.

Note on the issue's premise: #6258 cites a +27 net lines growth-guard violation from #6166. That premise is stale — the merged #6166 left onboard.ts net-smaller, and #6276/#6332 shrank it further, so the guard (.github/workflows/codebase-growth-guardrails.yaml, a per-PR net-neutral diff gate) is not currently red. This PR is therefore incremental maintainability work; it keeps onboard.ts net-neutral-or-smaller so the gate stays green.

Related Issue

Refs #6258

Changes

Add src/lib/onboard/created-sandbox-failure.ts: reportSandboxCreateFailure (warns-and-continues on an incomplete create; otherwise prints diagnostics + recovery hints and exits) and reportSandboxReadinessFailure (prints the readiness failure, defers cleanup to the Docker-GPU patch or deletes the failed sandbox, then exits). onboard.ts replaces the two inline blocks with module calls — net −2 lines.
Add src/lib/onboard/sandbox-create-step.ts: runSandboxCreateStep encapsulates the BuildKit prebuild handoff → Docker-GPU create-patch provisioning → create-stream behind a context + injected-deps boundary. This move is line-neutral on onboard.ts; its value is a named, unit-testable boundary (the prepare → patch → stream sequence is now testable without standing up the entrypoint), not a size reduction. Build-context and exit-listener cleanup stay with the caller that armed them.
Add focused unit tests across created-sandbox-failure.test.ts, sandbox-create-step.test.ts, sandbox-create-launch.test.ts, create-stream.test.ts, create-stream-argv.test.ts, and create-stream-ready-gate.test.ts covering failure branches, redaction, exit-code preservation, GPU vs non-GPU readiness cleanup, prebuild/patch/stream wiring, direct argv spawn boundaries, terminal-agent/default-driver ready-check gating, fail-closed onPoll error handling, and redacted poll/readiness trace behavior.

Type of Change

Code change (feature, bug fix, or refactor)
Code change with doc updates
Doc only (prose changes, no code sample modifications)
Doc only (includes code sample changes)

Quality Gates

Tests added or updated for changed behavior
Existing tests cover changed behavior — justification:
Tests not applicable — justification:
Docs updated for user-facing behavior changes
Docs not applicable — justification: no CLI flags, commands, configuration, or documented user workflow changed. The user-visible differences are limited to safer failure-path diagnostics (credential redaction), more accurate readiness-failure exit status, and fail-closed create-poll error handling.
Sensitive paths changed (security, policy, credentials, preflight, onboarding, inference, runner, sandbox, or messaging)
Sensitive-path review completed or maintainer-approved waiver recorded — reviewer/approval link/justification: onboarding/sandbox path. Extraction boundaries were verified against the pre-extraction source; intentional safety refinements are explicitly covered by tests: create output is redacted before failure logging, readiness failure preserves a non-zero create-stream status instead of flattening to 1, Docker-GPU during-create polling is isolated from readiness detection via onPoll, escaping poll errors abort create with classified/generic failure text plus redacted trace emission, and ready-check exceptions emit redacted trace evidence without falsely forcing Ready. Requesting maintainer sensitive-path review.
Non-success, skipped, or missing CI check accepted by maintainer — check name, approval link, and follow-up issue:

Verification

PR description includes the DCO sign-off declaration and every commit appears as Verified in GitHub
Normal pre-commit, commit-msg, and pre-push hooks passed, or npm run check:diff passed when hooks were skipped or unavailable
Targeted behavior tests pass for the current change set, or tests are marked not applicable above — command/result: npx vitest run --project cli src/lib/actions/sandbox/snapshot.test.ts src/lib/onboard/created-sandbox-failure.test.ts src/lib/onboard/sandbox-create-launch.test.ts src/lib/onboard/sandbox-create-step.test.ts src/lib/sandbox/create-stream.test.ts src/lib/sandbox/create-stream-ready-gate.test.ts src/lib/sandbox/create-stream-argv.test.ts → 107 passed; npx tsc --noEmit --pretty false --project tsconfig.cli.json → clean; npm run test-size:check → passed; npx prek run --all-files --stage pre-commit --skip source-shape-test-budget --skip test-skills-yaml → passed; npm run test-conditionals:scan -- --top 25 → no new changed-file conditional failures; git diff --check → clean.
Applicable broad gate passed — npm test for broad runtime/test-harness changes; npm run check for repo-wide validation/coverage changes — command/result: Required live E2E run 29021019725 passed on exact head 82518f7e6f69c39d9bc04f63a365753aac7b1b6d: cloud-onboard, onboard-repair, onboard-resume, state-backup-restore, upgrade-stale-sandbox, and snapshot-commands all succeeded. PR CI checks also passed on the same head after rerunning flaky policy-channel-list.test.ts shard timeout (cli-test-shards (3)).
Quality Gates section completed with required justifications or waivers
No secrets, API keys, or credentials committed
npm run docs builds without warnings (doc changes only)
Doc pages follow the style guide (doc changes only)
New doc pages include SPDX header and frontmatter (new pages only)

Signed-off-by: Dongni Yang dongniy@nvidia.com

Summary by CodeRabbit

New Features
- Improved sandbox onboarding using a dedicated creation step that centralizes prebuild handoff, streaming create execution, GPU patch wiring, and readiness capture.
Bug Fixes
- More consistent, centralized handling for both create failures and readiness failures, including redacted output, clearer diagnostics, and reliable retry guidance.
- Safer cleanup on readiness failures to avoid same-name collisions, with correct behavior for GPU-enabled flows.
Tests
- Added coverage for create/readiness failure reporting, exit-code fallback, cleanup/command messaging, and early-detach behavior in create-stream.

…dule The sandbox create/readiness failure handling was inlined in the ~4,900-line src/lib/onboard.ts entrypoint. Move both terminal handlers behind a focused boundary in src/lib/onboard/created-sandbox-failure.ts, mirroring the injected- deps shape of created-sandbox-finalization.ts (#6332): reportSandboxCreateFailure warns-and-continues on an incomplete create but prints diagnostics + recovery hints and exits otherwise; reportSandboxReadinessFailure prints the readiness failure, defers cleanup to the Docker-GPU patch or deletes the failed sandbox, then exits. Behavior is unchanged (verified byte-for-byte); onboard.ts ends net-smaller, satisfying the codebase-growth guard without a waiver. First cohesive increment toward the create-orchestration extraction; the remaining prebuild/create/finalize wiring stays for follow-ups. Refs #6258 Signed-off-by: Dongni Yang <dongniy@nvidia.com>

Move the BuildKit prebuild handoff, Docker-GPU create-patch provisioning, and create-stream out of the src/lib/onboard.ts entrypoint into src/lib/onboard/sandbox-create-step.ts (runSandboxCreateStep), behind a context + injected-deps boundary. Behavior is unchanged (pure move, verified byte-for-byte). The move is line-neutral on onboard.ts, but it gives the create step a name and makes the prepare->patch->stream orchestration unit-testable without standing up the 4,900-line entrypoint. Build-context and exit-listener cleanup stay with the caller that armed them. Second cohesive increment for #6258. Refs #6258 Signed-off-by: Dongni Yang <dongniy@nvidia.com>

coderabbitai · 2026-07-08T05:54:14Z

Note

Reviews paused

It looks like this branch is under active development. To avoid overwhelming you with review comments due to an influx of new commits, CodeRabbit has automatically paused this review. You can configure this behavior by changing the reviews.auto_review.auto_pause_after_reviewed_commits setting.

Use the following commands to manage reviews:

@coderabbitai resume to resume automatic reviews.
@coderabbitai review to trigger a single review.

Use the checkboxes below for quick actions:

▶️ Resume reviews
🔍 Trigger review

📝 Walkthrough

Walkthrough

onboard.ts now delegates sandbox creation to runSandboxCreateStep and sandbox create/readiness failure handling to created-sandbox-failure.ts. New tests cover the extracted orchestration and reporting paths.

Changes

Onboard sandbox create/failure extraction

Layer / File(s)	Summary
Sandbox create step module and wiring `src/lib/onboard/sandbox-create-step.ts`, `src/lib/onboard/sandbox-create-step.test.ts`, `src/lib/onboard.ts`, `src/lib/sandbox/create-stream.test.ts`	`runSandboxCreateStep` resolves launch inputs, creates the Docker-GPU patch, streams creation with readiness hooks, and returns the create result plus patch handles; `onboard.ts` now calls it, and tests cover launch forwarding, readiness patching, terminal-agent output gating, and VM detachment when output patterns are empty.
Failure reporting module and wiring `src/lib/onboard/created-sandbox-failure.ts`, `src/lib/onboard/created-sandbox-failure.test.ts`, `src/lib/onboard.ts`	`reportSandboxCreateFailure` and `reportSandboxReadinessFailure` centralize create/readiness diagnostics, recovery hints, sandbox deletion, Docker-GPU-specific cleanup, and exit behavior; `onboard.ts` now delegates both failure paths, and tests cover incomplete vs hard create failures plus readiness cleanup branches.

Estimated code review effort: 4 (Complex) | ~45 minutes

Suggested labels: refactor, area: onboarding, area: sandbox, area: architecture

Suggested reviewers: jyaunches, brandonpelfrey

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 50.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title accurately summarizes the main refactor: extracting sandbox create-failure reporting and the create step into modules.

✨ Finishing Touches

📝 Generate docstrings

Create stacked PR
Commit on current branch

🧪 Generate unit tests (beta)

Create PR with unit tests
Commit unit tests in branch fix/6258-extract-onboard-create-orchestration

_{Comment @coderabbitai help to get the list of available commands.}

github-code-quality · 2026-07-08T05:54:17Z

Code Coverage Overview

Languages: TypeScript

TypeScript / code-coverage/plugin

The overall coverage remains at 96%, unchanged from the main branch.

TypeScript / code-coverage/cli

The overall coverage in the fix/6258-extract-onb... branch remains at 77%, unchanged from the main branch.

Show a code coverage summary of the most impacted files.

File	main `0d2aca4`	fix/6258-extract-onb... `82518f7`	+/-
`src/lib/actions...ge-preflight.ts`	89%	74%	-15%
`src/lib/inference/config.ts`	98%	93%	-5%
`src/lib/sandbox...reate-stream.ts`	99%	99%	0%
`src/lib/onboard.ts`	28%	29%	+1%
`src/lib/actions...box/snapshot.ts`	79%	80%	+1%
`src/lib/onboard...dbox-failure.ts`	0%	100%	+100%
`src/lib/sandbox...est-fixtures.ts`	0%	100%	+100%
`src/lib/onboard...-create-step.ts`	0%	100%	+100%
`src/lib/sandbox...m-ready-gate.ts`	0%	100%	+100%
`src/lib/sandbox...eam-progress.ts`	0%	100%	+100%

_{Updated July 09, 2026 14:18 UTC
Code Coverage is in Public Preview. Learn more and provide us with your feedback.}

github-actions · 2026-07-08T05:56:51Z

PR Review Advisor — Informational

Merge posture: Informational / low confidence
Primary next action: Resolve or justify PRA-1: PR review advisor unavailable.
Open items: 0 required · 1 warning · 0 suggestions · 4 test follow-ups
Top item: PR review advisor unavailable

Action checklist

PRA-1 Resolve or justify: PR review advisor unavailable
PRA-T1 Add or justify test follow-up: Runtime validation
PRA-T2 Add or justify test follow-up: Runtime validation
PRA-T3 Add or justify test follow-up: Runtime validation
PRA-T4 Add or justify test follow-up: Runtime validation

Findings index

ID	Severity	Category	Location	Required action
`PRA-1`	Resolve/justify	correctness	—	Re-run the PR Review Advisor or perform a manual review.

Review findings by urgency: 0 required fixes, 1 item to resolve/justify, 0 in-scope improvements

⚠️ Resolve or justify before merge

Investigate these in the current review; either fix them, explain why they are not applicable, or document the accepted risk.

`PRA-1` Resolve/justify — PR review advisor unavailable

Location: not file-specific
Category: correctness
Problem: The automated advisor could not complete: PR review advisor SDK execution failed: session: security-trust must call pr_review_update_ledger exactly once (observed 2 starts); security-trust must finish pr_review_update_ledger successfully exactly once (observed 1 successful of 2 total completions); turn: security-trust: security-trust must call pr_review_update_ledger exactly once (observed 2 starts); security-trust must finish pr_review_update_ledger successfully exactly once (observed 1 successful of 2 total completions)
Impact: Automated review evidence is incomplete, so human review must cover the changed code manually.
Recommended action: Re-run the PR Review Advisor or perform a manual review.
Expected follow-up: Resolve in this PR or explain why the risk is acceptable.
Verification: Inspect the workflow logs and raw advisor artifact for the execution failure.
Missing regression test: No regression test recommendation is available because the advisor did not complete.
Done when: The risk is fixed or explicitly justified in the PR. Verification: Inspect the workflow logs and raw advisor artifact for the execution failure.
Evidence: PR review advisor SDK execution failed: session: security-trust must call pr_review_update_ledger exactly once (observed 2 starts); security-trust must finish pr_review_update_ledger successfully exactly once (observed 1 successful of 2 total completions); turn: security-trust: security-trust must call pr_review_update_ledger exactly once (observed 2 starts); security-trust must finish pr_review_update_ledger successfully exactly once (observed 1 successful of 2 total completions)

💡 In-scope improvements

These are lower-risk, not throwaway. Prefer fixing them in this PR when they are local to changed code; defer only with rationale or a linked follow-up.

None.

Test follow-ups to resolve or justify

If these cover changed behavior, prefer adding them in this PR; otherwise state why existing coverage is enough or link the follow-up.

PRA-T1 Runtime validation — Run the `onboard-repair` E2E job for Onboarding and sandbox state must converge across persisted metadata, reported status, and the live runtime. Matched files: `src/lib/actions/sandbox/snapshot-create-stream-test-types.ts`, `src/lib/actions/sandbox/snapshot.ts`, `src/lib/onboard.ts`, `src/lib/onboard/created-sandbox-failure.ts`, `src/lib/onboard/sandbox-create-launch.ts`.. Deterministic regression risks require live validation: lifecycle-state, upgrade-rebuild.
PRA-T2 Runtime validation — Run the `onboard-resume` E2E job for Onboarding and sandbox state must converge across persisted metadata, reported status, and the live runtime. Matched files: `src/lib/actions/sandbox/snapshot-create-stream-test-types.ts`, `src/lib/actions/sandbox/snapshot.ts`, `src/lib/onboard.ts`, `src/lib/onboard/created-sandbox-failure.ts`, `src/lib/onboard/sandbox-create-launch.ts`.. Deterministic regression risks require live validation: lifecycle-state, upgrade-rebuild.
PRA-T3 Runtime validation — Run the `state-backup-restore` E2E job for Upgrade, rebuild, snapshot, and restore operations must preserve user state while replacing stale runtime state. Matched files: `src/lib/actions/sandbox/snapshot-create-stream-test-types.ts`, `src/lib/actions/sandbox/snapshot.ts`.. Deterministic regression risks require live validation: lifecycle-state, upgrade-rebuild.
PRA-T4 Runtime validation — Run the `upgrade-stale-sandbox` E2E job for Upgrade, rebuild, snapshot, and restore operations must preserve user state while replacing stale runtime state. Matched files: `src/lib/actions/sandbox/snapshot-create-stream-test-types.ts`, `src/lib/actions/sandbox/snapshot.ts`.. Deterministic regression risks require live validation: lifecycle-state, upgrade-rebuild.

Workflow run details

This is an automated, non-binding review; it still expects maintainers and agents to respond to each required or warning item. Treat suggestions as current-PR improvements when they touch changed code; defer only with maintainer rationale or a linked follow-up. A human maintainer must make the final merge decision.

github-actions · 2026-07-08T05:57:05Z

PR Review Advisor (Nemotron Ultra) — Informational

Merge posture: Informational / low confidence
Primary next action: Resolve or justify PRA-1: PR review advisor unavailable.
Open items: 0 required · 1 warning · 0 suggestions · 4 test follow-ups
Top item: PR review advisor unavailable

Action checklist

PRA-1 Resolve or justify: PR review advisor unavailable
PRA-T1 Add or justify test follow-up: Runtime validation
PRA-T2 Add or justify test follow-up: Runtime validation
PRA-T3 Add or justify test follow-up: Runtime validation
PRA-T4 Add or justify test follow-up: Runtime validation

Findings index

ID	Severity	Category	Location	Required action
`PRA-1`	Resolve/justify	correctness	—	Re-run the PR Review Advisor or perform a manual review.

Review findings by urgency: 0 required fixes, 1 item to resolve/justify, 0 in-scope improvements

⚠️ Resolve or justify before merge

Investigate these in the current review; either fix them, explain why they are not applicable, or document the accepted risk.

`PRA-1` Resolve/justify — PR review advisor unavailable

Location: not file-specific
Category: correctness
Problem: The automated advisor could not complete: PR review advisor SDK execution failed: session: scope-risk-map called pr_review_update_ledger before analysis; turn: scope-risk-map: scope-risk-map called pr_review_update_ledger before analysis
Impact: Automated review evidence is incomplete, so human review must cover the changed code manually.
Recommended action: Re-run the PR Review Advisor or perform a manual review.
Expected follow-up: Resolve in this PR or explain why the risk is acceptable.
Verification: Inspect the workflow logs and raw advisor artifact for the execution failure.
Missing regression test: No regression test recommendation is available because the advisor did not complete.
Done when: The risk is fixed or explicitly justified in the PR. Verification: Inspect the workflow logs and raw advisor artifact for the execution failure.
Evidence: PR review advisor SDK execution failed: session: scope-risk-map called pr_review_update_ledger before analysis; turn: scope-risk-map: scope-risk-map called pr_review_update_ledger before analysis

💡 In-scope improvements

These are lower-risk, not throwaway. Prefer fixing them in this PR when they are local to changed code; defer only with rationale or a linked follow-up.

None.

Test follow-ups to resolve or justify

If these cover changed behavior, prefer adding them in this PR; otherwise state why existing coverage is enough or link the follow-up.

PRA-T1 Runtime validation — Run the `onboard-repair` E2E job for Onboarding and sandbox state must converge across persisted metadata, reported status, and the live runtime. Matched files: `src/lib/actions/sandbox/snapshot-create-stream-test-types.ts`, `src/lib/actions/sandbox/snapshot.ts`, `src/lib/onboard.ts`, `src/lib/onboard/created-sandbox-failure.ts`, `src/lib/onboard/sandbox-create-launch.ts`.. Deterministic regression risks require live validation: lifecycle-state, upgrade-rebuild.
PRA-T2 Runtime validation — Run the `onboard-resume` E2E job for Onboarding and sandbox state must converge across persisted metadata, reported status, and the live runtime. Matched files: `src/lib/actions/sandbox/snapshot-create-stream-test-types.ts`, `src/lib/actions/sandbox/snapshot.ts`, `src/lib/onboard.ts`, `src/lib/onboard/created-sandbox-failure.ts`, `src/lib/onboard/sandbox-create-launch.ts`.. Deterministic regression risks require live validation: lifecycle-state, upgrade-rebuild.
PRA-T3 Runtime validation — Run the `state-backup-restore` E2E job for Upgrade, rebuild, snapshot, and restore operations must preserve user state while replacing stale runtime state. Matched files: `src/lib/actions/sandbox/snapshot-create-stream-test-types.ts`, `src/lib/actions/sandbox/snapshot.ts`.. Deterministic regression risks require live validation: lifecycle-state, upgrade-rebuild.
PRA-T4 Runtime validation — Run the `upgrade-stale-sandbox` E2E job for Upgrade, rebuild, snapshot, and restore operations must preserve user state while replacing stale runtime state. Matched files: `src/lib/actions/sandbox/snapshot-create-stream-test-types.ts`, `src/lib/actions/sandbox/snapshot.ts`.. Deterministic regression risks require live validation: lifecycle-state, upgrade-rebuild.

Workflow run details

This is an automated, non-binding review; it still expects maintainers and agents to respond to each required or warning item. Treat suggestions as current-PR improvements when they touch changed code; defer only with maintainer rationale or a linked follow-up. A human maintainer must make the final merge decision.

github-actions · 2026-07-08T05:58:08Z

E2E Advisor Recommendation

Required E2E: onboard-repair, onboard-resume, state-backup-restore, upgrade-stale-sandbox, cloud-onboard
Optional E2E: sandbox-operations

Dispatch hint: onboard-repair,onboard-resume,state-backup-restore,upgrade-stale-sandbox,cloud-onboard

Workflow run

Full advisor summary

E2E Recommendation Advisor

Base: origin/main
Head: HEAD
Confidence: high

Required E2E

onboard-repair: Required by the deterministic lifecycle-state risk plan. The onboarding create/readiness/failure changes must be validated against repair flows where persisted metadata, gateway state, and live sandbox state must converge after partial failures.
onboard-resume: Required by the deterministic lifecycle-state risk plan. The sandbox create orchestration and failure reporting changes can affect resume after interrupted or incomplete onboarding, so live resume coverage is required rather than relying only on unit tests.
state-backup-restore: Required by the deterministic upgrade-rebuild risk plan. Snapshot restore auto-create and create-stream changes can affect backup restore correctness, retryability, and avoiding ghost destination sandboxes.
upgrade-stale-sandbox: Required by the deterministic upgrade-rebuild risk plan. Restore/rebuild flows must preserve user state while replacing stale runtime state, and this PR changes snapshot restore create execution and failure behavior.
cloud-onboard: The PR changes the full hosted onboarding sandbox creation path, including command argv construction, readiness gating, create failure handling, and cleanup after readiness failure. A full hosted onboarding run is required to validate the real user flow end to end.

Optional E2E

sandbox-operations: Useful adjacent coverage for general sandbox lifecycle operations because create-stream argv/readiness behavior and snapshot restore auto-create touch sandbox lifecycle mechanics.

New E2E recommendations

None.

Dispatch hint

Workflow: .github/workflows/e2e.yaml
jobs input: onboard-repair,onboard-resume,state-backup-restore,upgrade-stale-sandbox,cloud-onboard

github-actions · 2026-07-08T05:58:09Z

E2E Target Recommendation

Required E2E targets: onboard-repair, onboard-resume, state-backup-restore, upgrade-stale-sandbox
Optional E2E targets: None

Dispatch required E2E targets:

gh workflow run e2e.yaml --ref <pr-head-ref> --field jobs=onboard-repair
gh workflow run e2e.yaml --ref <pr-head-ref> --field jobs=onboard-resume
gh workflow run e2e.yaml --ref <pr-head-ref> --field jobs=state-backup-restore
gh workflow run e2e.yaml --ref <pr-head-ref> --field jobs=upgrade-stale-sandbox

Workflow run

Full E2E target advisor summary

E2E Target Advisor

Base: origin/main
Head: HEAD
Confidence: high

Required E2E targets

onboard-repair: Onboarding and sandbox state must converge across persisted metadata, reported status, and the live runtime.
- Dispatch: gh workflow run e2e.yaml --ref <pr-head-ref> --field jobs=onboard-repair
onboard-resume: Onboarding and sandbox state must converge across persisted metadata, reported status, and the live runtime.
- Dispatch: gh workflow run e2e.yaml --ref <pr-head-ref> --field jobs=onboard-resume
state-backup-restore: Upgrade, rebuild, snapshot, and restore operations must preserve user state while replacing stale runtime state.
- Dispatch: gh workflow run e2e.yaml --ref <pr-head-ref> --field jobs=state-backup-restore
upgrade-stale-sandbox: Upgrade, rebuild, snapshot, and restore operations must preserve user state while replacing stale runtime state.
- Dispatch: gh workflow run e2e.yaml --ref <pr-head-ref> --field jobs=upgrade-stale-sandbox

Optional E2E targets

None.