Robert Haas <robertmhaas@gmail.com> writes:
> So we have an established precedent that it is right to warn about
> things that are sketchy at the time that they are defined, but not
> every time they are used.
Sure, but we don't have that option available to us here --- or more
accurately, ALTER USER/DATABASE SET *does* warn if the search_path value
looks like it might be invalid according to the current context, but
that helps little for this problem. What's important is whether the
value is valid when we attempt to apply it.
Basically, I don't think you've made a strong case for changing this
behavior; nor have you explained what you think we should do instead.
regards, tom lane