Content aware analysing footage is taking forever

Content-Aware Fill is a resource hog. There are a few ways around the problem.

The comp that you make to use Content-Aware Fill should only contain one layer to start. When CAF is done only the fill layer and any reference frames should be in the comp. For safety, and to speed up the rest of your compositing tasks you should also render the CAF comp to a visually lossless production format like GoPro Cineform, ProRez, or even an image sequence and use that in the rest of your project. 

First, pre-edit your footage so you are not wasting any time processing frames that are not going to be in the final production.

Second, make sure you are not using a higher frame rate than you are going to use in the final production. If you shot at 60 and your final edit is 24 then drop your edited 60 fps footage in a 24 fps comp, then pre-compose the footage moving all attributes to the new comp, and run CAF on the nested comp.

If the frame size is over HD and the area you are trying to fill is only about 1/4 of the frame, duplicate the footage layer, pre-compose the duplicate, open the pre-comp and use the region of interest to crop the comp to just the area that you need to fix plus enough extra space for the fill to work, pre-compose again, and run CAF on the smaller clip. This will prevent CAF from running calculations on a 4K frame when you are only trying to fill a hole that is about 1/8 of the frame.

The biggest mistake I see people make with CAF is to fail to trim the clip to only the frames that are needed. I run into the same thing all the time with Camera Tracking, Warp Stabilizing, and Rotobrush and roto work. Just yesterday I helped someone out with a Rotobrush project where they were trying to remove most of an actor's body from a shot that was nearly 12 seconds long when all that was needed was a mask for the actor's head and neck for about 30 frames. The solution was to duplicate the layer, set a new in and out point for the layer, draw a layer mask a little larger than the actor's head, and run the process on the critical 30 frames. Those 30 frames were then used as a track matte to fix the 30 frames that spoiled the shot. 

I hope this helps. If you have any more questions we need workflow, design, and shot details so we can make specific recommendations for your shot.