If the problem is related to an organizational concern or design problem my go to pattern is to stop and wait for 8 to 12 hours. By that point several competing concerns have surfaced and I am in a much safer place for making arbitrary decisions that carry massive second and third order consequences.
I immerse myself in the problem domain. It'll sound odd, but I try to become the domain within which the problem arise. Then I turn the problem around and around until I find a better fit. Kind of like 3D tetris