Investigating...

I've failed to reproduce this - do you have the workflow you saw it in?

These tasks were probably both internally queued and runahead limited at the same time which might possibly be the cause of the issue?

Pretty sure that's not supposed to happen. Tasks should get queued only on release from runahead.

do you have the workflow you saw it in?

Yes, but I don't think it will help much (it's a beast).

Here's a GUI screenshot showing a running task with a running job that is in the runahead state:

If I see this in the wild I will try to grab some graph GraphQL output and test in Tui to check whether the issue is definitely data store side.

These tasks were probably both internally queued and runahead limited at the same time which might possibly be the cause of the issue?

Pretty sure that's not supposed to happen.

This was more of a "steps to reproduce" than a description of what happened at the implementation level. I meant that the task could be considered to be held back by both runahead limit and (likely multiple) queues, perhaps the interaction between the two could be the root cause.

(might be fixed by recent changes: #4906 ?)

This hasn't been seen in the wild for a while so closing this issue. Please re-open if this is spotted again.