I think Terry Pratchett once supplied the appropriate response to this type of advice:

“ARGHHHHH NO NO NO NO NO!”

Even thinking about writing SQL to write SQL to drop anything that is invalid is a bad idea. (And the supplied code has an error anyway which means it can drop a public synonym because an identically named private synonym has become invalid.)

I’ve added a couple of comments on the feedback suggesting that the note should be removed, or significantly modified.

I agree – more significantly there ought to be some clear and timely documentation about which bits of functionality are massively dependent. Watch out, for example, for the XMLDB owner – if I recall correctly you can drop this one without problems, but expdp and impdp (datapump) will stop working.

I think in this case you can even go through the DBCA install process and cross XML off the selected install list – and then find that datapump doesn’t work.

(Someone may want to correct me on that – I may be wrong.)

I’m not sure they could afford me ;)

But what I do is a lot easier than what they often have to do. I started by seeing a carefully prepared 45 minute presentation about what went wrong and how to limit the damaged; and then I had some free time to play with the problem and no pressure to get a result or respond to other calls.

I think Coskan may have given you some of the possible arguments. In the case of the presenter it related a mechanical procedure for upgrading their product at client sites. (A procedure that had worked perfectly well for quite a long time)

Thanks for the comment on my presentation. I have to say that I thought it went very well (even though I skipped the second example). It’s a presentation that I will be repeating at various venues around the world over the next nine months.

Warning: Although my fix worked for me, I can’t guarantee it would work for everyone else. It’s just possible that unexpected things like “operation X happened between the drop and the recreate” may make my method fail.

I guess no one ever thought that actual relevant objects may become invalid and stay that way until they are referenced. I notice on one test db it would drop apex and some app views that are granted to public.

With moderation like that, who needs guns?

And I thought the old notes about index rebuilds were bad, you could rewrite that note as
Step 1: Point Gun at head
Step 2: Pull trigger