My approach is that on my main FX tracks, there's never an RTAS running at all. Stages don't take well to this stuff usually., especially when they realize a bunch of your plugins won't necessarily load or load correctly.
That said, I build in layers (where appropriate, of course) and try my best to region group parts of a sound catorgically (e.g. for a gun combat segment I might do: grab, pistol whip, gun mvt (for pistol whip), gun mvt (handling), gun cock, shot, shot wide, shell eject, etc) or by frequency (e.g. for a sting, I might do: hi, mid, lo, tonal). Laid out from a top-down approach and strategically organized toward what the mixer is probably going to want (e.g. same recurring sounds on same track, consistantly start a multi-track region group on the Odd numbered FX track, etc) That's for the final FX tracks.
I also carry many mono and stereo 'source' tracks which serve as a sandbox. Usually it's where I build my edits and comp them down, which are then pulled into the FX tracks. A great example as a vehicle engine. The engine, in one case I was working on, consisted of about 8 mic angles per take for the vehicle source itself, and they had to be weaved together to create a 'Start, double tap rev, very hard rev out' maneuver. Some angles on the car were muted depending on the segment I was using (or because some mics were not good on some parts), but overall the raw edit of this vehicle move was about 25 tracks wide (in order to carry all mic POVs that I needed and still checkerboard the edit). Then I would go through and comp down the entire maneuver into a mono' Engine' and a 'Tailpipe' POV, which means that what ends up on my final FX track is 2 mono files of this engine. That's all the stage want's and needs, especially if the Tailpipe can be provided separate from the Engine - sometimes, that's not the case with some vehicles because there is not discrete Tailpipe POV on the source, or the POV exists but blows (no pun intended), and the stage is thus better using EQ on it. Once these mono's are generated, I'll do split outs and gain-adjust for POV cuts that happen - so that way, the stage isn't committed to how the POVs are and essentially have handles on the edit tracks - so ultimately what we did here was use 'source' to create 'new source' that we edit with. I rinse and repeat for other vehicle items like tires, skids/peels, suspension, cornering, brakes, shifter - each comp'd down to mono tracks stacking under the engine in the FX tracks (and POV split from there as needed).
I use the same technique on other things too, like doors - unless it's some sort of elaborate tech door or vault door, I limit myself to 3-4 elements. If the door's any wider, I'll comp down appropriately complimenting elements to bring it down to 3-4. Again this is sort of thing in the mentality of the stage and what the mixer wants (e.g. enough, but not too much).
In these cases, I keep the region-grouped 'source' edits muted on the source tracks and always carry these tracks in my edit. In the event the stage calls because they need to dig into a sound elements deeper, then it's easy to ship off the source edits themselves. Over time and practice of using proper discrimination and confidence in your edits, this stage situation should usually never happen because you'll kept comp'd elements appropriately separated from others which you know the mixer will need discrete control over (hence, comping by frequency range or categorically). But you still never know - so I always keep these source edits.
Sometimes I have RTAS plugins in them to automate specific processes on the fly, especially when an effect is non-linear in nature (e.g. speed ramp), especially when doing design so I can create proper sonic textures and gestures to match what's happening. But even then, those get comp'd down and pulled up into the FX tracks (in this case, Design tracks sitting below my FX tracks) to serve as anew 'source' element to build into the overall design composite which gets region grouped.
Aside from all that, I usually just AS a source element with processing and commit it to the file if I even process a file (e.g. EQ, ring modulation, etc). I prefer keeping my edits layered and raw, and commit anything which actually requires processing. In my opinion, nearly no amount of processing can make up for using the wrong source elements to create a sound. The only real exception to that is when your (a) creating something conceptual like sound design (b) correcting for a problem with the source element (e.g. low end rumble or getting rid of the muddy bass, a steady tone needing to be notched out, etc). If your having to constant EQ and process everything before it hits the FX tracks (hence, running all the processing) then it may be wise to go back and make soure you're working with the right source elements and not trying to cajole bad/poor ones into getting the job done.
Hope it helps! That's my 2 cents at least.