I think it's more important to think about what state of mind are you in - for example, if I'm tired, I'm just going to go home a lot of the time. I know that it isn't an efficient use of my time if I keep banging on a problem while not in a state to solve it efficiently.

This is something many academics learn quickly, as it is important in order to solve deep problems, especially with the workload that is levied on academics.

Myself, I've had many instances where this has happened, including with open source work.