Hey,
My experience is that you shouldn't normalise to 0db, but rather a little bit off like -1db. If let's say you burn your track to a cd, and your software/burner accidentally introduces an extra bit during the burn, you could get a pop. There is also what Architect said, that if there is no headroom it makes it difficult to do any addition processing on the file without it clipping.
However, there is a BIG difference between compressing a file and bringing it up to 0dB, and simply normalising up to 0dB...if you compress the audio, you are effectively changing the dynamics destuctively which cannot be undone later on if needed. Normalising simply brings up all the levels so that the highest value is at 0dB and everything else is brought up equally.
Unless there's something I don't understand about normalising, any engineer worth his salt should be able to rip the track into a wave editor and normalise to a lower value before adding any more processing, this should not damage the file at all at it is only removing bits that were artificially introduced when you normalised it.
Take care,
Nick