Affinity is actually a good thing, as it avoids invalidating core L1 and L2 caches during a context switch. Those cache invalidates and loads actually waste CPU cycles, hence why in HPC-land coders spend an inordinate amount of time trying to maximise affinity.
MSFS 2020 is multi-threaded but it is not embarrassingly-parallel. It appears (to me anyway) that there is always one main thread doing more work than all the others. Moving that thread around on the cores causes L1/L2 cache invalidates so you’re actually harming performance (albeit at a very low level that is unlikely to translate to noticeable changes in FPS). That thread will burn X CPU cycles on whatever core it runs, you can’t speed it up by turning off affinity. All you do is see even core CPU usage but the total CPU it burns in any wall-clock interval is still X.
For HPC applications however, shaving even 0.1% off of run-time by maximising affinity can actually translate to many hours of CPU time being saved (and wall clock time as well). Which means researchers get their results sooner and your department saves a few $$$ of HPC compute cost.
Not a criticism, just an FYI 