Even in sciences where it shouldn't be the norm, we tend to have "dumb" people reliant on smart computers for interpretation of results, instead of smart people using dumb computers as tools to aid computation.
Frankly, even if you still believe there's reason that certain research scientists would need that kind of processing power and storage at times, that doesn't excuse the fact that the OS should not be taking up that time. An OS should exist to load user programs, allow data access by users and programs, provide a platform for peripheral connections as needed, and otherwise get out of the way with a bare minimum load.
Most tools within the sciences that require vast computation power aren't pretty, nor output data in a nice, readable form. What do you think university supercomputers are for? They run extremely stripped-down operating systems[0] designed for a single application & single user.
I'm at a research university (as a student) and I guarantee that most home-grown tools here are not user-friendly, but exist to get the job done. This data is then interpreted in other, more user-friendly programs/environments like Tableu/Matlab/Julia/whatever.
BS. Applications in scientific research easily blow that ridiculous limit out of the water. Your point about single vs multi user is good, but you do yourself no favors by starting off with a bad premise. And lest you say you're talking about personal OS only, that's not what you wrote at all.