I would say Doom 3 is probably not a good idea for Inf's next engine. It is probably able to render outdoor environments better than most people give it credit for; according to Raven Quake 4 will actually have some pretty big outdoor areas with lots of enemies and not much slow down, but outdoor is clearly not what the D3 engine was intended for. Given that Inf is pretty big on outdoor areas it doesn't make sense to go to an engine that will create trouble when most of the advantages of the D3 engine probably won't be usefull in Inf. Real time shadows might be a nice touch in certain moments of CQB, but I doubt it would make up for the limitations in other areas.
At the same time, Inf's next iteration may not have such an emphasis on outdoor areas that the Far Cry engine is a necessity. FC can render areas that are miles and miles long, but how likely is it that Inf will have multi-mile maps? I'm sure the Inf player base would grow in the transition to a shiny new engine, but with super-mega-huge maps even a 32-player match would be slower than 1v1 on Extreme Prejudice. If the FC engine proves prone to bugs and/or the lack of support becomes a problem, it may not be worth the trouble given Inf maps probably won't be of the super-behemoth size that is the main advantage of the FC engine.
If all else fails, the UT2004 engine would do just fine. It is still capable of doing large outdoor areas 5x bigger than any current Inf map and look damn good in the process with very little slowdown. Plus, modding for UT2004 is made to be as easy to mod as possible. I don't want to discount the advantages of other engines, but if the complications therein prove too great, UT2004 seems like the reliable, flexible 'safe bet' engine.