What webhook retry backoff schedule does Jira Cloud apply before permanent failure?

Context

Jira Cloud delivers webhooks with an exponential-backoff retry policy documented only at a high level: retries continue for up to 14 days. The exact backoff intervals, jitter implementation, and the criteria that classify a delivery as permanently failed are not published.

Goal

Determine whether an integration can rely on deterministic retry timing for idempotency-key design, dead-letter queue sizing, and alerting thresholds when testing against a developer instance.

Constraints and uncertainty

• Developer instances enforce a 100 req/min rate limit, which may truncate or distort the production retry curve.
• Forge tunnel sessions expire after two hours, preventing observation of multi-day retry sequences in a single local run.
• WireMock can simulate 429 responses with Retry-After headers, but it cannot replicate Atlassian's undocumented backoff schedule or the conditions that trigger permanent failure.

Questions

1. What are the exact backoff intervals and jitter bounds Jira Cloud uses between retry attempts?
2. Which HTTP response codes or timeout conditions cause Jira to mark a webhook as permanently failed before the 14-day window elapses?
3. Does the retry behavior differ between developer instances and production tenants in a way that invalidates local retry-test results?