P: "lrcat-data" and "lrcat" files should be merged into one

"Why 2 files which both store photo metadata, are kept separate?"

The .lrcat file is a SQLite relational database that stores the myriad "facts" about a photo -- metadata fields, develop settings, the location in the file system, etc. Each "fact" typically ranges in size from a byte or two to several hundred bytes. If the the .lrcat file gets lost without a backup or without .xmp files, you're totally screwed, losing all the information about your photos.

The .lrcat-data file is a RocksDB persistent data store, storing the actual bit masks computed for AI masks and Remove. These bit masks can be tens of thousands of bytes each. If the .lrcat-data file gets deleted, LR will recompute all the bit masks, which could take a long time, and in theory, the recomputation should produce the same exact results.

Relational databases like SQLite are designed for  storing millions of little facts.  Whereas the RocksDB is designed for storing larger blobs of bytes more efficiently.  I presume that Adobe chose to split out the masks into the RocksDB .lrcata-data file because they believed it would yield significant performance advantages. I haven't seen any concrete evidence one way or the other.

Well, read about that file here:

• https://helpx.adobe.com/lightroom-classic/kb/catalog-faq-lightroom.html

(And to infuriate you more notice that it can be included in the backup IF it is large enough)

Note:

    In case of images with an AI mask, yourcatalogname.lrcat-data file will be > 0 K to "In case of images with an AI mask or AI Edits, yourcatalogname.lrcat-data file will be > 0 K".
    If the file is larger than 0 KB, the .lrcat and .lrcat-data will be included in the zipped backup
    In case the file is missing or is at 0 Kb, the .lrcat is included in the zipper backup

And P.S. see:

• https://fstoppers.com/business/youre-probably-not-backing-all-your-lightroom-classic-data-526597

(yes that link is a bit dated)