I see that it's faster when I profile kate/kdev, but I cannot easily write a benchmark for this. It's only noticeable when the KDE_COLOR_SCHEME_PATH variable is set, otherwise the global application config will be used afte rall, which is going to be shared most probably. Any idea how I could construct a valid KDE_COLOR_SCHEME_PATH path to e.g. the breeze color scheme from within a kcolor scheme auto test?

I also noticed that on application startup two default color schemes are created: One by plasma-integration set on QApplication and the other by QStyle::standardPalette() in the widget style.
I couldn't figure out a way to avoid this (have QStyle use the default app palette set by p-i or something without dirty hacks like using dynamic properties to communicate that we already have created the proper palette elsewhere). Perhaps this is something you could also profile/investigate :)

sure, but first let's get this in. @broulik or @dfaure care to give your +1?

pushed this now with a proper benchmark too, shows a ~10x performance win when a non-empty PATH is set