If I install a package while it is in [testing] and then it gets moved to [core], which repository should it report?

Good motivating example, another one might be [community] -> [extra] (pkgver/pkgrel may have remained the same while the PKGBUILD changes). In your example, I think -Q should report testing, -S should report core (since it's displaying info for the package in the repo db, not in the installed pkgs db).

But then if I "pacman -S <pkg>" again, it will now report [core]. Exactly the same package and files, different metadata.

Perhaps if you gave an example of what you are trying to achieve, I can understand what information is needed.

I have multiple repositories that can provide the same package with the same pkgver/pkgrel, and need to know (e.g. in a puppet module, in my cfg mgmt, etc.) what the installed version is. This is especially important when migrating a host from production<->staging repositories, for example. It's really exactly as described: I want to know where the installed package came from, not just what it's name and version are. The packages might differ between the repositories (locally applied patches, remotely differing PKGBUILDs but same pkgver/pkgrels, etc.). The installed package metadata should carry the repository name that it was downloaded/installed from when cached in the pacman local db, since it carries pretty much everything else already.

For context: I have production (frozen-in-time with security and performance patches only, 6 monthly update cycle), staging (testing for production-next), and upstream mirror (aka archlinux) package repositories/dbs. This is for a deployment of a bunch of ArchLinux-based machines for which just following Arch repos directly would be hazardous to uptime.

I think that you should be using a different version to signify different package versions. You could also use the "packager" field if that was not suitable.

I am fairly sure we have rejected a similar request before. I'll leave this open, but it is very low on the priority list.

If I did it myself and submitted a patch, would that make it more likely to happen sooner? (I won't have time for about a month to do this myself, it's on my low priority wishlist.)

Yes - but I'd strongly advise discussion of the feature on the pacman-dev mailing list first to judge whether it would be accepted.

I'd also suggest that it is only shown as additional information in -Qii output.