Bug: cloud sandbox resume returns 409 when checkpoint is not yet available

[devin-ai-integration]

Bug: cloud sandbox resume returns 409 when checkpoint is not yet available

Description

When a sandbox is suspended, attempting to resume it with agentuity cloud sandbox resume can return a 409 Conflict error if the checkpoint has not yet been fully created/persisted.

Steps to Reproduce

1. Have a running sandbox that gets suspended
2. Immediately attempt to resume it:

   agentuity cloud sandbox resume sbx_<id> --org-id <org_id>
   

3. Observe a 409 error response:

   sessionId: 'sess_8a205f6491595ffb85ae1a54413bca86',
   status: 409,
   url: 'https://catalyst-usc.agentuity.cloud/sandbox/sbx_558180614bc2b3b361d940aefcf...'
   

Expected Behavior

The resume command should either:

• Wait/retry until the checkpoint is available before attempting to resume, or
• Return a clear, descriptive error message indicating the checkpoint is not yet ready (e.g., "Sandbox checkpoint is still being created. Please try again shortly."), or
• Implement a polling mechanism that waits for the checkpoint to become available before issuing the resume request

Actual Behavior

A raw 409 status code is returned with no indication that this is a transient/timing issue. The user has no way to know they should simply wait and retry.

Additional Context

• This appears to be a race condition: after a sandbox is suspended, it takes some time for the checkpoint to be fully persisted. If the resume is attempted before the checkpoint exists, the API returns 409.
• Trying another sandbox that had been suspended for longer worked fine, confirming it's a timing issue.
• The fix could be applied at the CLI level (retry logic with backoff), the API level (queue the resume until checkpoint is ready), or both.

Environment

• CLI command: agentuity cloud sandbox resume
• Catalyst endpoint: catalyst-usc.agentuity.cloud

[linear]

AGENT-1455