How to prevent faulty worker from consuming all flows and putting them in error state

Help
1

I have a pool of ~6 workers each on their own machine, and sometimes one will have a hardware error. When this happens, it can quickly consume all flows and put them in an error state.

How can I have the flow go back into the work queue for an attempt by another worker in this case?

2

I’d be open to any creative solution, I may just be fundamentally thinking about prefect flows incorrectly!