: They all could be converted, in theory, but the first 3 would definitely be.
I haven't done any programming in quite a while and never 3D modeling, so I would defer to you as to implementation. Of course we would all like the coolest of effects, but backwards compatibility and stability would be my top criteria. I would also stress that the ability to create and edit maps on all platforms is a key. Although there is a growing PC contingent playing Marathon and 3rd party variants, thanks to the influx of interest from Halo, the core of the community is Mac-centric. Creating a new game engine that requires a PC to fully take advantage of its features would permanently split the community along platform lines - we can't afford that kind of fragmentation.
One other thought - I know that I was the first person to question what you're smoking when you proposed creating a new game engine from scratch. You obviously have the skills and the drive to make it happen. I still wonder if you're barking up the wrong tree, however. There already is a good effort to port Marathon to the UT engine, and a smaller effort to port it to Quake 3. There are in fact excellent game engines already out there that can do everything you say and then some. The problem is that converting existing Marathon maps, shapes, physics and other files to work with these is exceptionally tedious - it basically means starting from scratch. Even worse is the fact that the only available editing tools are on the PC, precluding the bulk of existing Marathon mapmakers from participation. I think that perhaps a better use of your talents would be to create a Mac or cross-platform (eg. X11) universal map editor. This editor would be able to create new levels from scratch that could be used with the more advanced game engines (at least one of them in any case) and it would have the ability to import existing Marathon and AO maps, shape files etc. directly such that they could be used immediately with the newer game engines. This tool would be far more valuable than a new engine, and it will be needed anyway even if you build a completely new game engine.
