What it is
A CPU limit is a budget per period, for example 100 ms of CPU time every 100 ms. Several threads can spend that budget in a fraction of the period, and then every thread in the group waits for the next period. This lab runs a small model of that mechanism, together with memory.max and memory.high reclaim, OOM kills, pids.max and pressure stall information, so you can change a file and watch the counters and the latency.
When to use it
Use it to see why a burst of requests gets slower under a limit although average CPU is low, why cpu.weight only matters when groups compete, why page cache is reclaimed before a process is killed, why memory.high slows allocation without killing, and why threads count against pids.max.
Gotchas & limits
The lab models mechanism and relative size, not real-machine performance. It has no golden data from a real cgroup v2 host yet; per-CPU 5 ms slices, a fair-share model between sibling groups and a simplified memory.high penalty are approximations. There is no host memory pressure, so memory.min and memory.low have no effect, and cpu.max.burst, cpuset and I/O control are not modelled.