To avoid the thundering herd problem when retrying is triggered at exactly the same moment by many entities allow for passing in a jitter maximum value which will be used to randomly alter the amount of time slept by each retrying object user; this helps randomize the thundering herd and makes it possible to reduce the issue.
When we stop retrying because of recurring exceptions, we expect the decorator to raise the last exception. (if wrap_exception==False) This commit fixes issue #8.
…opriate regression tests. (fixes #4)
…frames, adjust api and tests #1
…rarchy could slip through, preparing for 1.0.1 release
…Error instead of just raising it directly, add a few more comments