l/''xx'' templates
So, what should we do with these l/xx templates? I think they are no longer preferred over l.
I don't know. Can you see if the pages in Category:Proto-Slavic appendices are not slower when used with {{l}} instead?
I've already tested that (for different languages and scripts). For a page with 1000 linking templates it took almost 12s for old version of l, almost 7s for Luacized l (current version), and almost 5s for l/xx to be loaded. In practice, the number of l templates are usually much less then that, so I think the difference is not significant or, at best, is not enough to outweighs the disadvantages.
The disadvantages are that the features and templates wouldn't be centralized if we continue using them, and that l/xx has much less features than l.
The lack of features is the point, though. That's why they are fast. They are optimised for one purpose and nothing else. If you really want to get rid of those templates, I think you need to convince more people than just me. Grease pit would be a good place to go.
I know, but I don't want them to be removed/deprecated at the moment, but to remove that section that you added in l's docs which says "l/xx is strongly preferred".
I suppose if {{l}} is now faster than it is before, it is not as strongly preferred anymore.