No, it’s about in-game speed. When all mods are enabled (I have lot of mods I didn’t or cannot publish yet), all slows.
In 0.92 it was worse, in 0.92b speed increased.
gotar,
The question is: is some speedup worth such efforts?
I can either spend 1-2 weeks to add zip support and get easy to use system.
Or I can spend 1-2 months to write all new code and new tools, explain to everyone that they have to use these tools (and btw - why you forgot about “zips are hiding files from user” - because that’s what custom formats do). Get a headache because we found a user with platform that is neither unix nor Win and needs custom code for mapped files.
I heavily doubt that this will give us big improvements over current behavior.
Interesting - I didn’t knew that situation is changing that fast. Previously most of posts here had WinXP useragent while now we have more Linux users than WinXP 
Probably this as well - current json validation is not as fast as I wanted it to be. Will fix this after release.
It also validates everything every time. This is really slow. Lazy validation would remove this problem but only if there will be some way to detect changes - like checksum.
I don’t know what may affect this. More mods will result in larger memory usage but should not have any effect on CPU. Most “heavy” areas like adventure map or battles load everything on start so drive access time should not have any effect either.
I don’t know current codebase so I cannot say, what effort is required. I was thinking about converting OH3 data into raws and trying to load these in my way, but first of all I don’t know lod format and would have to use VCMI code for that and we’re back to square one.
Maybe some day we’ll come back to this discussion. If there’s going to be central mod repo, converting them in future shall not be a big problem.