Codespaces cannot run local jobs in the background for more than 60m

[bloodearnest]

Just surfacing a reported user issue with codespaces, in case its not already known - you can't leave large projects running and come back when they're done, as the codespace will shutdown after 60m of no UI interactivity. AFAICT, there's no way to keep a codespace alive except for UI interaction.

The reporting user was using local development to workaround this, as it could run overnight.

Increasing the timeout may work, but increase costs. Perhaps it could be increased if a specific user requests it? The max seems to be 240min (4h).

Alternatively, having the jobs run in some other compute location could work, but that's complex and challenging.

Perhaps the recommended solution is to use local development?

[bloodearnest]

This UX issue interacts with opensafely-core/opensafely-cli#377, as restarting a codespace that has been suspend but was running jobs at the time leaves the local openasfely run state potentially confused, making it hard for the use to resume where they left off.

[lucyb]

Removing this from the board as we don't intend to do anything on this right now, but I'm keeping it open as I think we could provide better guidance and processes around this.