What it is
When a service slows down, clients time out and retry, and the retries add load to a service that was already struggling. This sandbox sends identical traffic through several client and server strategies, injects a fault, and measures what each one does to goodput, load on the service, wasted work and time to recover.
When to use it
Use it to see why backoff alone does not stop a retry storm, what a retry budget or circuit breaker changes, and why a long queue can turn a short slowdown into a long outage. Compare strategies, then replay one on a timeline and inject faults while it runs.
Gotchas & limits
This is a teaching model with one service and one client population. Defaults are illustrative and are not the values of any library or cloud service; real systems add connection pools, load balancing and many hops. The numbers show a mechanism, not a prediction. Recovery time needs a few seconds of healthy traffic before the fault to have a baseline.