It's sometimes mentioned that unitary, coherent, noise can adversely impact the performance of QEC codes compared to purely stochastic Pauli noise. This motivates protocols such as randomized compiling which tailors noise to stochastic Pauli errors via Pauli twirling. On the other hand, it's a well known property of quantum error correction that repeated syndrome measurements decoheres unitary errors. So how exactly worse are coherent errors for QEC compared to incoherent errors?
