We have limited information but so far Afterburner appears to be only a decode accelerator for ProRes and ProRes RAW. I actually question the need for ProRes acceleration since that codec is already so lightweight, but maybe at 8k it helps.
It seems unlikely to accelerate FX. Those are often GPU-based and an FPGA cannot out-perform a GPU ASIC. However there is a slim possibility that some unique task (e.g. video noise reduction) which does not map well to either CPU or GPU methods might contain underlying primitives which could be accelerated by an FPGA.
Since the FPGA is reprogrammable it's possible additional functions could be added in the future - provided those are within the performance envelope of an FPGA. E.g, it could accelerate H264 encode/decode. We know that because academic researchers have already done those and published the results.
However the Mac Pro already has full-custom (ie ASIC) H264/H265 encode/decode acceleration on the T2 chip. There would seem little benefit to using a slower FPGA implementation for this.
If some new compute-intensive codec not handled by T2 later become widely used, in that case Afterburner might be programmed to accelerate this.
Re accelerating export of unrendered timelines, that task consists of executing the stored list of edit decisions and then encoding those in the selected codec. It is a varying mix of CPU and GPU actions. In general I don't think an FPGA could greatly improve this.
However if code paths for common render and export tasks were profiled, and if some "hot spot" was found which could not be optimized via normal CPU or GPU methods, then hypothetically an FPGA might accelerate this. But that's a lot of "ifs".
We will know in due time exactly what Afterburner brings to the table. The new Mac Pro will be released and many independent benchmarks will be run.