@Carl Love Yes you probably want to emphasize it has a computer science related causality before someone starts a new field of math called catasrtrophe theory.
Ok i will get around at some point to looking at all of the case examples for which i have archived that the erroraneous output is recieved or remains unevaluated.
It happens equally as often for the floor function as it does for frac in my experience, and is mostly trancendental functions. Sometimes it occurs just for particular expressions that involve integer powers of pi and rational coefficients, a while ago i did post a way that i was able to deterministically express the the errors in evalf using certain interpolations of the curve fitting function taking the number of digits as a variable, it will be some where in the history of questions on this forum no doubt.
Most are able to be fixed it really isn't major bug at all, just alot more frequently occuring may have previously been anticipated by those responsible for the development of the floating point approximation inbuilt code, I guess a combination of me being quite naive with respect to the inner workings of maple and also quite pedantic might explain why i keep coming across them.
Once I have spent a reasonable amount of time looking at the cumulative content i will post them here.