As I thought it might, this does not duplicate on my machine.
If you look into the code, the comparison that is failing is tested twice, once to low precision and once to high precision. Only the high-precision test is failing. This makes me suspect that it is a numerical noise error that is not exposed on any of the architectures we normally test on (and therefore not a problem). I think you’re okay to ignore it; if you want to try to dig into the C++ we can.