Comment by jeffbee

Just limiting your retry budget to 1% of normal rates using a client-local token bucket with no distributed coordination will eliminate the possibility of long-lived retry storms.