handle failed removals

[Simbiat]

practice showed that in some cases removal of instances can fail (in my case because of deadlocks). This is fine when removal is done manually, since there will be output to the user, but if it fails in automation with one-time jobs, it becomes a problem, since a one-time job may get ran twice, which may have side-effects. the only approach I can thing of right now is to have a separate table like failed to remove and exclude all instances that are there, and clean those as part of unhanging process, but not sure how good this is. It is probably not urgent, though, since really one-time job should be deigned in a way, that will not do damage if ran multiple times anyway.