I think it is contextual, but it's a useful approach in many cases: it's effectively a sunk cost vs value of information analysis: if you've already sunk most of the cost into something and there's a problem which means it isn't going to work, then the potentially valuable information from learning about the next steps is probably pretty cheap, and reduces the risk that the investment in the next attempt isn't wasted. On the other hand, if you know it's not going to work and the next steps involve a lot of extra time/material investment but aren't going to tell you much, then it's probably a good idea to throw the current attempt away and try again.