The current implementation of retry for Bitbucket rate limits sleeps 5 seconds and tries again.
We have Bitbucket folders that have hundreds of repos some of which have hundreds of branches. This can sometimes take hours to do a folder scan. During that time the Bitbucket rate limit is constantly at its maximum
Adding exponential backoff (with jitter for randomness) will allow other API requests (such as to download the Jenkinsfile when a multibranch pipeline is building) to go through and will alleviate the constant pressure on the rate limit.