Seems a bit weird to me, the operator is supposed to run after dependency graph update. The only case when it's violated in the operator redo. But that's not what the T58073 and T55494 are about?
I think this was a mistake. I realized shortly after i created this diff that it actually did not fix what i thought it would fix :( and i should have deleted it right away, sorry.
@Gaia Clary (gaiaclary), happens. Consider joining discussion/review in T55494! :)