With Rational[1, 6] the NProbability gives 0, with N@Rational[1, 6] the NProbability gives 0.90. In either case, 500 is not a big number. If there were numerical computation issues (tags: workingprecision, machineprecision, etcprecision), the NProbability should have problems with N@Rational but not with Rational, because Rational is exact. You're calling it a bug, me too.
The person at "Wolfram Technical Support" said that it is not a bug. And closed the case. Meaning that he is not going to forward it to the development team for review. Noice.
He filtered me right at the gate. No chance of getting this issue through to the programmers.