In my proposed use, there is no real need for the timing to be closely synched as each chip is going to be handling differing data and outputting independantly anyways.
Most applications with multiple micro-controllers don't need it, but there could be certain applications where it would be desireable or necessary (for instance, a multiple micro-controller parallel processing implementation where you need things in tight sync for the code to work properly between/across the processors, or perhaps where you have two processors on the same address bus sharing memory - those kinds of apps).
In that kind of a situation, you need an external clock circuit, and there are other issues related to timing and clock signal trace lengths that can come into play (though less likely at the slower speed of the ATMega - but it is still possible, depending on the distance between the processors on the PCB and a whole host of other reasons). This is an area where things can get arcane and touchy; one of the "magic" areas of electronics and computer design (generally in electronics, when things start moving into the analog realm - which of course all electronics is based on mainly - it starts to almost become an art instead of a science - magic!).
