Hi,
Here I am again, presenting another dilemma to the forum in the hope of getting some clarity from those in the know! 
I would like to start by stating (as I have before) that the After Effects fourm might not be the best place for this question, but it is the only place I know to start with - all the help I have received here over the past year makes me want to post all my questions here.. (Ill keep the "Why is the sky blue type questions for google though!
)
Ok, Here we go.
We are now shooting on a Red Scarlet. All footage is being shot at 4K resolution - Mostly for post movement and zooming options, and the rest, for the options it gives in post - planned or not. We are running 4 machines at the office, 2 MAC's and 2 PC's. The PC I work on has the best specs, the second PC matches the first, with the exception of the screen card which is one version behind (and which is not supported by CS6). The two MAC's are not worth mentioning now 
I am trying to figure out the best workflow, starting at R3D RAWfootage, down the line to MXF OP1a deliverables. An easy question I am sure! 
I am aware that there is a lot to still work out, but I need to get the starting questions sorted, and thats where I hope the forum can assist. We had a DIT come to the office for a day to give advice, and he ran us through a great workflow, however, I am not sure it is best suited for what we are trying to do - am I correct in saying that no workflow is universal, almost every project may require tweaks on its specific workflow? Here is the one suggested to us:
1) Shoot 4K R3D RAW
2) Open in RedCineX and transcode to HD Editorials files (quicktime, Avid DNxHD codec, Debayer set to (screenshot below))
3) Import into Premier and perform first cut edit (offlines) - save EDL from this cut
4) Import EDL into RedCineX and export new masters from original R3D RAW files as OpenEXR or DPX sequences.
5) Import new masters back into Premier dynamic link to After Effects to complete composits (effects etc)
6) Media encoder to export various final renders
In the above workflow, the last two steps were added by me, as the workflow stopped at step 4. This looks great, but because of our project needs, I am not sure how to get this to work. Keeping in mind, almost all shots have green screen miniature characters and they require retiming with twixtor and stabalizing etc for the addition of the Lip Synch tool we use. I am also really not sure how to work with EDL files, or even what they are used for - I realize this is a big gap in my lnowledge, but there we go - I hate to admit that I dont even understand timecode, or what exactly it would be needed for. I am starting to get to grips with it, and so far understand that timecode is important for being able to recover from data loss of footage - and getting quickly back to the point where the loss occured by knowing what part of what clip to recover. I am also concerned, that exporting these lower resolution intermediated wont help us, as they come out at 1920X1080, so even in the rough edit, I cant do the post moves we want to. By looking at this, I need to work on all shots at 4K resolution - So the work between Premier and After Effects starts getting blurry here. Which do I do first, cut in Premier, with rough keying and rough cropping to allow the layers to all be visible and then go to After Effects for cleaning up, Or do i start in after effects, Key, retime and then go to premier for the cut. The problem with this approach, is I then have no way of knowing which parts of the clip to extract and retime, as I have not seen the edit to know how well the layers will interact.
Whew, I am getting myself in a bundle here, and I am hoping that I am uneccesarily complicating everything! 
I have been running this through from many angles, and watching many videos on Red Workflows etc, and come up with the following (and then my questions and concerns about it afterwards). Lets set the scene - 2 Characters, one background set. 6 second clip. Background seen, Character one and two come into shot from each side of the screen and move to the centre - full body shot.(2 sec). Characters meet in the middle and camera zooms to closer shot of heads (done with manipulating the 4K images) 1 sec. zooom back out to original shot (1 sec). Both characters move to the right and camera pans along background to follow them (1 sec) Movement done by manipulating 4K background plate. Now this is a simple scene, and I am using it to test a workfow that will make sense. Characters are shot against green screen, set is shot with locked off camera, shot with the post movement in mind. Each footage clip is a full ten second shot to allow for selection of best movement.
1) Shoot 4K R3D RAW
2) Open in RedCineX and trancode to highly compressed 4096X2160 intermediate codec (still to figure out which one)
3) I originally thought to open in Prelude and ingest only selected part of the footage, but it has been pointed out to me that we need to see the characters interactingto know what parts of the 10 second clip we would need (and I was just starting to get to know prelude!)
4) Open editorial clips in Premier and perform a rough edit. Rough cropping and keying and retiming with twixtor (as needed) to see characters interact against background.
5) Replace with an After Effects comp with dynamic link. Perform keying, retiming (as needed) and any other effects (including lip synch and facial manipulations with Liquify tool)
6) Back in Premier, confirm that scene is reacting and looking as expected.
7) In After Effects, import original R3D RAW footage and simply replace the video layers as needed. This way, we have been able to perform most of the work, with reasonable performance from the PC.
8) Re Key (I assume this would be needed if the footage is replaced) but all other manipulations should remain intact?
9) Collect project save, save premier project and send them on to colour team to be graded.
Does this even make sense?
Say I do the following
1) Shoot 4K R3D RAW
2) Open in RedCineX, perform a light grade, correcting ISO and exposure to get a flat look accross all three pieces of footage for the scene.
3) Transcode and export as a DPX or OpenExr sequence and never look back to the R3D RAW. Will the image sequence still contain enough original image information to be of use to the colourist? Will the light grade be burned in with the sequence?
4) Use new masters and follow the steps above to complete the scene.
I apologize for the extremely long post, if nothing else, I hope it provided a laugh for a few (thinking, Poor chap...he has NO IDEA!). 
I also apologize for placing this in the AE forum, but there we go.
Any advice would be appreciated, any thoughts as well. Just in closing, I did get this clip day before yesterday http://tv.adobe.com/watch/learn-after-effects-cs6/editing-with-red-ado be-premiere-pro-edit-up-to-5k-with-ease/ And it was great, I tested out using 4K RAW throughout the workflow, and it does work when the resolution is dropped, however, due to the precision of the face manipulations, I dont think it is a viable solution with the hardware we have.
Well, there you have it, the ramblings of the mad man end here for the day - thank you for taking the time to read them..
Pierre