If I recall correctly, this would manifest as timeouts following a spurious HTTP 409 Conflict: We understand that we are pushing some number of bytes, but have failed to rewind the file handle to the beginning after the first request; therefore, the handle has fewer bytes to push than expected/expressed in the request headers, the server should keep the connection open (expecting the indicated number of bytes to arrive) eventually killing the request with a timeout.
No reliable mechanism known to expose; however, seems to be something which has haunted Fedora for a long time (at least via the REST API)... Our original Tuque code is largely an adaptation of previously existing code from the Python "fcrepo" library
Lacking reproduction steps, seems to fall down to regression testing; making sure valid PUT requests succeed. In particular here, datastream modifications, which perform "PUT" requests with a body.
discoverygarden inc. | Managing Digital Content