After runnig a development instance in Tasks, everything died

[bartekgatzcoder]

I have stumbled upon the following problem.

• Opened a GitHub issue: Tasks prompt and tasks list are not separated well enough coder#20423
• Added a "traiage" label to start the GitHub integration to Coder Tasks
• The Agentic AI reported it did the thing
• I asked Agent to run a development instance of Coder inside of the workspace (because there is no "preview" application configured for this template)
• I saw a bunch of notifications about a number of ports being forwarded
• I decided to check the "workspace view" to get a better idea on how to open the http ports forwarded from the workspace (tried 8080, but it returned an error)
• everything stopped reacting, including the complete task view
• after a few minutes, the workspace became alive again, but the developer instance was gone

[johnstcn]

Currently, the template you used will run tests automatically after the agent enters the "idle" status.

My guess is that the workspace suffered some kind of temporary resource exhaustion from running make test and develop.sh at the same time.

If you can reproduce the issue, I can investigate removing the automatic test run behaviour from the template and see if that resolves the issue.

[bartekgatzcoder]

This may have been the thing. The tests reported "failure" at the same time the workspace came back to life.

[johnstcn]

We now use 'write coder on coder' by default for the workflow.