Looking through the errata for the MPC5777m - 0N50N, I noticed e8731 that states received data may be shifted by one bit, or sampled off by one - I tried to do some testing to observe this at different baud rates and clock speeds, but wasn't able to trigger it. I also turned on the suggested 0 parity and parity error check, but haven't had it trip a frame error yet (only 1 day testing so far).
How often does this happen in practice? Is it more common in some configurations than others? Has any one observed it in the wild? Would always using the 16x oversample mode help?