The problem
Should a cursor update, a transaction, and background indexing be treated as equally important just because each is runnable?
What current research shows
The scheduler provides mechanisms for priorities, cgroups, utilization clamps, deadlines, and policy classes. Those mechanisms let user space express requirements, but the meaning of business importance isn’t available from CPU time accounting alone.
Where the evidence stops
Priority isn’t a magic solution. Bad priorities can starve work, increase energy use, or make the system less predictable.
What Valen Systems is testing
Test whether a declared work objective plus bounded priority policy improves completion of valuable work while preserving interactive and background service guarantees.