With CUPS_FORMATS_CACHE_OPT=1, an in-memory cache is enabled that stores the MIME type lists determined by mimeFilter() for each printer model (make_model); this allows subsequent queues with the same model to directly adopt the formats already found instead of performing the costly filter discovery again.
Optional add-on ENV:CUPS_CANONICAL_DEST

Please let me know, what you think about.
Thx.

I need to set aside some time to look at this, and I have my doubts that this duplicate PPD configuration is all that common.

That said, I'm thinking we could just cache based on the cupsFilter lines in each PPD file vs. hashing the PPD itself. That would offer the same benefits across all PPDs (regardless of whether they are the same) - all PS printers that handle "application/vnd.cups-postscript" will report the same supported document formats, for example. If a printer supports multiple media types (multiple cupsFilter lines) then we can either cache based on the set of supported media types (e.g. "image/urf+image/jpeg" for an AirPrint printer) or merge the cached lists of source types that can be filtered (that should be a little faster but I'm not sure whether the optimization would be worthwhile...)

Thank you so much for sharing your thoughts. I just wanted to mention that I don't assume this PR will be accepted as is.
I totally agree with you that the root cause triggering this investigation is not very common, but it is for me at least!-)
I really appreciate your suggestion, and I'll try to address the cupsFilter instead of the entire PPD.

Thx.