We have a report of access_tokens not being fetched due to failed connection. The user indicates that the request should be automatically retried in this circumstance. Can you implement a retry mechanism with an incremental backoff while fetching access_tokens?