MMCSS, gives options for controlling your audio system. Like CPU affinity and different priority settings, that you can not access by any other means. Now, program affinity and priority is something different. I have had great success in the past fiddling around with these parameters in endless different configurations. The ultimate goal, finding that one combination that lets your system fly, not one single pop or click, no matter what you do, going as low on latency as possible, this is for live usage. You could also call this a quest for stability.
One technical aspect of this would be to keep processing threads on a fixed core, using the same cache, avoiding them flying around. It can also be used to fin-grain control what cores do what, this is very important to spread load over all the cores and never overload a singel core. This also helps audio synchronizations between cores so you do not wast to much resources on this.
My setup is one daw per core and I use totalmix for syncing and mixing. This takes away lots of stress from the cpu. Because you basically off-load the digital synchronization of audio streams to the rme hardware, (almost like analog).
So without the MMCSS system, I have no way to control where the audio system is doing the processing... This can lead to some pops and click under unfortunate situations... Even if latency monitor is green for hours.
And as a kick to RME, this is why Linux would be a much more flexible and powerful option then running 50 year old windows system on modern hardware to get good results. But this is like talking to the hand. Per this day, no real audio operating system exists. One did and that was Be OS, but that got butchered.