I am not sure, if this case is similiar to bug 16441, but on [1] the null revision for the protect action is based on the wrong base revision (because protection should have a 0 byte diff).
There is a difference from 1 second between the save before the protect, so this can also be a slave lag.
For the same reason the diff hint on [2] is wrong, because it tells you, that there are 0 users between the two diffs.

[1] https://de.wikipedia.org/w/index.php?title=Zuckerrohr&diff=119540067&oldid=119540066
[2] https://de.wikipedia.org/w/index.php?title=Zuckerrohr&diff=119540067&oldid=119540062

Needs testing, if the null revision is loaded from slave.

Version: 1.22.0
Severity: normal