How does the mechanism of distributing incoming network interrupts among CPU cores (RSS, Receive Side Scaling) in Linux work? What happens if all network packets from Nginx start being processed by only one CPU core — core #0 — and how to fix this?