File hash check doesn't take errors into account

[ThrawnCA]

The hash check for skipping XLoader only takes the previous successful run into account, not failed runs, even though failed runs can still wipe the datastore.

This means that if you upload a valid file, then an invalid one that causes parsing errors, then try to re-upload the valid one, the file will be skipped because the hash is unchanged, and the datastore will be left unpopulated.

[duttonw]

There is a workaround for this by using the manual submit on the resource. This does mean there is a major time delay if missed.