Backup chain corrupted due to scp failure

[Bouke]

Today I've been trying to salvage some lost data using our duplicity backups. However during the restoration I've run into the following error a few times:

    Invalid data - SHA1 hash mismatch for file:
    duplicity-inc.20180601T033002Z.to.20180601T073003Z.vol16.difftar.gpg

Digging through the backup volumes, that specific increment caught my attention:

    200M Jun 1 09:38 duplicity-inc.20180601T033002Z.to.20180601T073003Z.vol15.difftar.gpg
     21M Jun 1 09:38 duplicity-inc.20180601T033002Z.to.20180601T073003Z.vol16.difftar.gpg
    201M Jun 1 10:27 duplicity-inc.20180601T033002Z.to.20180601T073003Z.vol19.difftar.gpg

Notably:

 - volume 16 is incomplete
 - volume 17 and 18 are missing

In the logs when creating that backup, the following relevant messages have been outputted:

    Giving up after 5 attempts. BackendException: Error running 'scp -oServerAliveInterval=15 -oServerAliveCountMax=2 /tmp/duplicity-VA2x4D-tempdir/mktemp-fqSlsK-18 offsite-backup.cb.local:/backup/dbs02r2.cb.local/duplicity-inc.20180601T033002Z.to.20180601T073003Z.vol16.difftar.gpg'
    Giving up after 5 attempts. BackendException: Error running 'scp -oServerAliveInterval=15 -oServerAliveCountMax=2 /tmp/duplicity-VA2x4D-tempdir/mktemp-RbVN9m-19 offsite-backup.cb.local:/backup/dbs02r2.cb.local/duplicity-inc.20180601T033002Z.to.20180601T073003Z.vol17.difftar.gpg'
    Giving up after 5 attempts. BackendException: Error running 'scp -oServerAliveInterval=15 -oServerAliveCountMax=2 /tmp/duplicity-VA2x4D-tempdir/mktemp-SEdTr8-20 offsite-backup.cb.local:/backup/dbs02r2.cb.local/duplicity-inc.20180601T033002Z.to.20180601T073003Z.vol18.difftar.gpg'

For reference, the command used to produce this backup is:

    /usr/bin/duplicity --full-if-older-than 1M --encrypt-key XXXXXXXX --max-blocksize 1048576 --gpg-options -z 1 --asynchronous-upload -v 8 --include /my/data --exclude ** /my pexpect+scp://offsite-backup//backup

Duplicity version: 0.7.15
Python version: 2.7.5
OS: CentOS 7.3 EL7 (centos-release-7-3.1611.el7.centos.x86_64)
Filesystem: XFS

Looking through the backup logs, I see quite a few of those "giving up" messages. Probably those increments are corrupted as well. I'm starting to lose my confidence in the backups produced by duplicity. I expect that when an increment hasn't fully completed, duplicity would not continue to use that increment for the backup chain.

Am I doing something horribly wrong, or is this expected behaviour from duplicity?