From it seems it is possible for an in-progress Synapse multi part upload state to be inconsistent with an S3 multi part upload state which is either completed or aborted. As a result any upload attempt will fail, e.g.
We should catch this case and automatically reset the upload with a forceRetry.
Originally observed on the below but not specific to the below and will reproduce through version 2.1.0
Python 2.7.10
synapseclient version is 1.9.4
Red Hat Enterprise Linux Server release 6.10 (Santiago)