> Exactly. The application is typically going to throw a "concurrent
> update" type of error when this happens, and we don't want magic
> background operations to cause that.
>
I`d give up the possibility of using CTIDs in the way You explained
for an auto-debloater without blinking an eye. Maybe we should have a
GUC to enable/disable the auto-debloater? Make it a reloption?