After migrating from OES2SP3 to OES2015SP1, my largest volume full backup (most of the time) stop progressing after about 800MB or about 40 minutes of elapsed time.
The process is still running, but is endless, the .sidf does not grow anymore, no messages on the source or destination server, it does not always stop in the same directory.
Sometimes the full backup works.
The full backup schedule has been changed : not better.
The backup has been split in two (.sidf create, then append) : not better.
Skipping some directories : not better.
TSAFS options have been reset to defaults : not better.

The full .sidf file is about 1TB, and there is plenty of available place.
The incremental backups works fine (monday to friday).
Other full backups of volumes work fine (weekend).

All the volumes are clustered.

Looks like it is more related to smdrd or tsafs, but no clue why it happens, any ideas are welcome.