The archive search will run on any page that might have discussions, this includes wgExtraSignatureNamespaces as well as talk namespaces. I suspect office and wikitech have the main namespace in that list.

As for non-existent pages, I suppose it is possible for a talk page to have been deleted but still have archives, although more likely there would be a redirect. Either way it's a rare edge case so I'm not too concerned.

@Esanders I vaguely recall a series of older DT tasks about the problems with DT using wgExtraSignatureNamespaces in this way to make a seemingly final decision that a page is a talk page when there are numerous indicators to the contrary, and/or requesting for ways to create a way to either opt-in or opt-out from that decision.

The main opt-in mechanism that I imagine would have been described there is __NEWSECTIONLINK__, which is clearly not set on any of these pages. Another (depending on the use case) is the complete lack of any comments/signatures currently on the page, although that's probably only enough signal to avoid creating reply payloads, not for creating new topics or finding archives (since some archive bots are configured to clear talk pages even if that means they become empty after that, sans for __NEWSECTIONLINK__ and whatever other header the page has).

I'm with this task being closed as a duplicate of a general task for "Avoid DT talk stuff on non-pages in signature-capable namespaces", if such task exists already, I imagine we won't need one for every DT feature, assuming that a solution will naturally piggyback off the same central switch.