I compiled 4 targets, the same dependent target got rebuilt 4 times. I don't think we have that much variance in our jar resolves, and if we did, you'd think it would have been a problem in 0.0.53.