In N=(2,2) Calabi-Yau GLSMs it should be possible to transfer B-branes between different phases, or to put it another way, two CY varieties which arise as different GIT quotients of the same space should be derived equivalent. For abelian gauge groups this is well-understood but the non-abelian case is still very hard. I’ll discuss the example of stratified Mukai flops which is well-known as a difficult test case. In recent work (with Wei Tseu) we found a simple geometric construction of the Fourier-Mukai kernel for the derived equivalence in this example.