the lighting model and the multiplayer modes are unrelated
what you describe is in essence light-lag
this kind of thing is neglegible
its a far cry from important compared to clipping or other gameplay calculations, when it comes to synchronizing for multiplayer
physics, etc, are all the same across clients, and can be predicted easilly.
regardless of accuracy, physics is something present in all games (realistic or not), and has been handled previously in a client/server model, and can be handled again in a C/S model.
more realistic technical limitations are :
" average pc sucks and wont run this smoothely with too many players "
" average net connection sucks too much to run this with too many players USING p2p "
and p2p systems have less lag in general than client/server models, so long as the node count is low.
hence, a low node count (4).
(C/S imagine everyone with 100 ping, that's 100 to the server, and 100 out to the next person, 200.
p2p, same situation, that's 100 to the next person.
half the latency.)
if everyone has to talk to everyone, your bandwidth required is (basic requirement * (number of players - 1))
so if you need 4 K a sec to sync with someone, thats 12 K a sec up for a 4 way game. thats about the limit of any 768/128 kb/s dsl connection
quake3 currently needs about 3K a sec to sync properly at minimum.
so you have some trade offs.
basically doom3 will have half the ping it would have had with a client server model, but will need more bandwidth.
note : p2p usually is designed in a way where it runs at the speed of the slowest node, so it may also suffer if you get an HPB on there, but maybe they did a work around for this. it would be obvious to do so, but historically people haven't, and it wouldn't be very unlikely for doom3 to have this issue.