You can (as far as I know) bridge an identical clock source between two different micros without any issues. It would probably be safer to use a CMOS clock source, because these are powered and have a powered output.
What I though of right as I was about to post this is that a crystal is inert. That is, it is a passive device. Power has to be applied across it to make it oscillate. I think it is unlikely, but what could possibly happen is if you fully bridge the crystal to both, it could cause one to see a DC signal, or have conflicting operations because the microcontroller is actually pumping a small, but measurable amount of power into the crystal to sustain its oscillations.
Might be worth a try on a breadboard first. I don't think that is very likely, and a single crystal should do. Namely because even if one sees DC temporarily, the common crystal will begin to resonate and prime both with the frequency AC signal.
The only thing this means is that the two ATmegas will operate synchronously, although rarely is this a bad thing.