In v1, we detected when users were engaging with workspaces to keep them alive.

This made for a great developer experience since the developer never had to think about "bumping". The workspace magically stayed on for however long they needed it. While this was great for developers, it was painful for those responsible for cloud costs. Tabs left open overnight and untraceable ghost activity beefed up cloud bills.

Organizations fall on both sides of this trade off.

I believe there's a way to get the best of both worlds here.

We could add a flag like coder config-ssh --keep-alive. Every tunnel command in the SSH config would carry along that --keep-alive flag, and every coder CLI spawned with the flag would periodically bump the workspace. Idempotent bumping (see #2230) is a necessary prerequisite.

Eventually we could bump within native IDE extensions. We could implement a similar keep-alive behavior on requests to coder_apps so browser customers get activity-based shutdown as well.

Implementation note: reuse the notification goroutine