Be easier to go with a processor with more memory.'1284 just as easy to work, similar in size to two '328's and the extra crystal & caps & resistor.
Could be, as long as a bootload file exists and it knows about fuse settings, etc.
You could run your 328s on the internal 8 MHz clock and save the crystals. You probably need 0.1 uF caps on the two VCC lines (VCC and AVCC). Apart from one set of 4.7 K pull-ups for I2C you should be able to get away with a pretty minimal setup.Make sure you give each slave a different I2C address, or alternatively do an I2C broadcast and talk to all at once, which might suit you better.