Symptom

All webhook retries fire within ~60ms of each other, hammering merchant endpoints instead of backing off. Merchants complain about being flooded with rapid retry attempts when their servers are recovering from an outage.

What you know

Dispatch uses exponential backoff for retries. The base delay is 1000ms (1 second), and the multiplier is 2x per attempt. The maximum delay should cap at 60 seconds.

Expected backoff schedule:

• Attempt 0: ~1,000ms (1s)
• Attempt 1: ~2,000ms (2s)
• Attempt 2: ~4,000ms (4s)
• Attempt 3: ~8,000ms (8s)
• Attempt 4: ~16,000ms (16s)

Actual behavior: every retry fires after ~60ms regardless of attempt number.

BASE_RETRY_DELAY  = 1000      ← milliseconds
MAX_RETRY_DELAY   = 60        ← what unit is this?
BACKOFF_MULTIPLIER = 2

[backoff-calculator] Backoff calculated
  attempt=0
  exponentialDelay=1000
  cappedDelay=60               ← capped to 60 instead of 1000?
  finalDelay=57

[backoff-calculator] Backoff calculated
  attempt=2
  exponentialDelay=4000
  cappedDelay=60               ← still 60
  finalDelay=63