COG database was last updated in 2020, which you probably know. Just for the sake of other readers:
As they indicate in the section titled "Expanded genome coverage", it was too computationally intensive to study all available genomes and MAGs - the majority of both bacteria and archaea in GTDB are MAGs. It was too demanding to study even complete genomes, so they settled on a smaller group. Given the number of COGs (4877), I think that group is representative of both kingdoms even though it is based on a relatively small fraction of the total.
There is some evidence of novel and unusual protein families in more recently discovered bacteria:
Still, it is a safe bet that many of those are just divergent variants of known proteins, as it is unlikely that these groups evolved thousands of protein families different from all other bacteria.
The first reference above has more ideas about doing this kind of analysis if you want to try it on your own, which I don't recommend. I have done it for about 5,000 MAGs, and it is very difficult to set up and takes a great deal of time, memory and general resources. If you are not easily dissuaded, this may help: