chore(api): log reason for failing to connect, resume sandboxes on state change

[matthewlouisbrockman]

currently, we drop the err from err = a.orchestrator.WaitForStateChange(ctx, teamID, sandboxID). This makes it hard for us to know why requests during pausing errored.

This change logs the dropped err so we can get better confidence in why we're gettign 500 errors during state transitions.

also updates the error logging to use the telemetry.reportCriticalError to maintain Error text while helping with the spans

---

Note

Low Risk
Low risk: changes are limited to additional telemetry/error reporting on existing failure paths, without altering core sandbox state logic.

Overview
Adds telemetry.ReportCriticalError calls in sandbox connect/resume handlers to capture the underlying errors (including WaitForStateChange failures, unexpected non-running/unknown states, snapshot fetch failures, and secure envd token errors) and to attach sandbox/team/build/template identifiers for easier debugging of 500s during state transitions.

^{Written by Cursor Bugbot for commit ab21f5a. This will update automatically on new commits. Configure here.}

[claude]

WaitForStateChange returns ctx.Err() when the polling loop context is cancelled (state_change.go:250). A normal client disconnect will trigger this path, logging a spurious Error and attempting to send a 500 to a client that is already gone.

Consider skipping the Error log when the error is a context cancellation or deadline exceeded.

[matthewlouisbrockman]

making 'em debug

[claude]

Same issue: WaitForStateChange returns ctx.Err() on context cancellation, so client disconnects will be logged at Error level. Filter context.Canceled / context.DeadlineExceeded before logging.

[matthewlouisbrockman]

making 'em debug, i want to know they're happening

[dobrac]

you can use the telemetry.ReportError and telemetry.ReportCriticalError instead to also populate the traces correctly. It does logging automatically

[matthewlouisbrockman]

you can use the telemetry.ReportError and telemetry.ReportCriticalError instead to also populate the traces correctly. It does logging automatically

shoud we add him to logger.L().Error(ctx, "Error getting last snapshot", logger.WithSandboxID(sandboxID), zap.Error(err)) too to debug those guys? (i think it's a 30s timeout somewhere but can't find it yet)

[matthewlouisbrockman]

eh, can worry about getting the reason for the failed db guys later, pretty sure those are all just timing out on the 30 second callback from the redis route

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}