Bug in validation procedure ?

A quorum requires at least three "valid" results in the sense that the results must be successful (i.e. not terminated with a client error) and syntactically correct (e.g. must consist of finite numbers). The results are then compared to find an "agreement" about a "canonical result". The results that are "close enough" to this canonical result are considered to be "valid" and the median of the claimed credits for the valid results is assigned to all of them.

I can't see a bug in the validation from the WU you are pointing to here. However there might be some further adjustment needed to what is considered to be "close enough" with the new AltiVec Mac App, whose results sometimes seem to be a bit futher away from the Windows mainstream.

BM