A client that always retries turns one slow dependency into an outage: each caller multiplies its load exactly when the dependency is least able to serve it. A retry budget caps the total retry rate (for example, no more than 10% of calls may be retries), which keeps the retry mechanism available for real transient failures without letting it amplify load during brownouts. In the session we compute the amplification by hand — a 50%-failure dependency with 2 retries becomes 3× traffic — and then set the budget and jitter policy together.
Study-clone note: the original session transcript and live-room embeds are not mirrored; this page reproduces the session layout with an original briefing.