With the change from #2635, completeImageBuild can now update the same build row twice in a transaction. This is something we've avoided in the past because it could case bad behavior in postgres (deadlocks).
completeImageBuild
I'm not sure if this is still something that we need to be careful about with postgres, but we should investigate to be sure.
I don't think that the double update should be required to fix #2634. It may be that there's a better fix anyway.
Metadata Update from @mikem: - Custom field Size adjusted to None
This issue has been migrated to Fedora Forge: https://forge.fedoraproject.org/koji/koji/issues/2658
Please continue any further discussion there.