So the trick to sanding longer with abrasives is wet sanding. In addition, in automotive work, a drop of Palmolive dish soap is added to a bucket of water. This addition makes a huge difference.

Overall, the principal of like polishes like is important. In abstract, polish is just fine abrasion. Like your finger prints are around 5k-7k grit equivalent. Rub something long enough and you will both polish and abrade it the same as this grit. The oils in your skin are the polishing agent.

I have played around with 10k grit wet sanding and then machine polishing with a light compound where places I rested my hand showed minor variations after stripping any oils and fillers with wax and grease remover (solvent).

I can think of several aspects to increase the complexity here. One could add inserts into the outer vibrating shell. These could be any materials.

I think the bigger issue will actually be the distance between the object and the shell. You see, the size of the random orbital action is the product of two concentric circles. In the pro automotive world, these are pneumatically driven. There are several models available with different properties related to this motion and the internal balance of the mechanism. Within this range of actuation, it is critical that abrasion does not follow a path of repetition. I think this likely means the shell must be larger than the radius of the largest of these two circles or maybe a more complicated size larger than the combination of overlapping radii including their central connection point. This should enable the part to move within the range of random sanding action. That range means the sanding is over a larger area.

The best shell is likely one with gaps similar to a DA sander with ports for dust collection.

Very little of any fiber touches the actual nozzle during printing. The actual fiber size used in filament is far far smaller than what most people imagine. It is only the waste dust from the production and processing of carbon fiber. All actual fibers of any useful length are sold in industry for use in composites. There are continuous fiber printers, but that is not at all related to what is used in 3d printing. If you actually look at the data from people testing materials, fiber infused materials are always weaker. They print better because they are breaking up the polymer bonds. Lots of people jump on the buzzword thinking it is technomagic mor betterer but do not pay attention to the details. If the fiber had any length to it, it would clog like crazy because a long bunch of fibers distributed in 1.75mm crammed into 0.4mm is never going to happen. It is just like a dust additive that happens to be available and is compatible. So it should be well distributed throughout. With ABS a wipe of acetone should help too, if left to completely flash off the solvent for a week or more. That needs to be super limited though. Acetone tends to get retained in bad bad ways with ABS. It is a massive no no to use in automotive applications.