I reran the "EinsteinToolkitReleased" Jenkins job today, and it has 4 failing tests.  The code has not changed since the previous run on 7th October (when all tests passed), but the build and test node is using a different OS and CPU.  The change in CPU was unavoidable as the old node is no longer available.  The change in OS was somewhat unintentional, but I suspect the problem comes from the CPU.  I believe that this was also observed for trunk; is that correct?  3 out of 4 failures are related to NaNs.  The 4th is not.