Let's say that you got unmatching CRCs for several tries using several rip modes and drives (EAC). Then you got a matching one in a later try. Can that be trusted as a perfect rip? I mean if the CRCs are different for each attempt, both the test and read CRCs change from attempt to attempt, it's just that the last try got it right, can you rely on it? Could it be that BOTH of them are not the correct one for the file that was intended when mastering, and that they both match? It's like getting a answer from a wrong source, and checking it against the same source.

Matching CRC's are not a guarantee that the rip was without error. It does make it more likely than not... But there is always the possibility of a consistent error.

Is that possibility really really small? Like 1/1000000000000?

By the way, this is for a scratched CD.

Is it possible that light is deflected in the same (wrong) way each time it reads?

It depends solely on the disc and the drive doing the ripping. Guessing at odds is pretty pointless.

http://en.wikipedia.org/wiki/Cross-interle...-Solomon_coding