Provide better merging primitives
About a decade ago, when people started releasing open fonts on a big scale, and the fontconfig main role changed from managing a scarcity of unicode-coverage limited fonts to managing an abundance of font files, Fedora defined with the then fontconfig maintainer the following template
https://pagure.io/fontpackages/blob/master/f/fontconfig-templates/merging-font-template.conf
It was a quick hack to manage the Droid font Google had just released, it didn't work too well (it needs to override font versions to work so it will break some version selection logic within fontconfig) and we all hoped it would not be necessary long term. Surely Google would soon stop this insanity of splitting a perfectly fine font family in multiple language subsets? (remember, we used fontconfig to construct synthetic font families out of unrelated font files, because creating wide-coverage fonts was hard, nobody even imagined someone would deliberately break up a wide-encoding font if this graal was ever achieved).
We know now, of course, that this splitting was deliberate, and would become common, because web devs want small font for their web sites (and to hell if they can not render all kinds of i18n text) embedded mobile people would rather use small country-specific fonts than large universal fonts that eat up their storage, and so on.
And besides things like Noto are just too heavy to be installed by default in full mode on today's desktops.
So fonts split by region or script are not a fact of life. And users will often not have on system all the font blocks corresponding to a single font family (from a design POW). And the region split changes over time because entities like Google hide the split in their UI, and optimize the split according to current market needs (this is very clear in AOSP's git history for example).
Therefore fontconfig should hide this technical complexity from users, present them just the font family design name, and do the juggling between all the bits of a font internally without requiring explicit manual selection of a specific font part.
Please provide a fontconfig configuration primitive that allows a system integrator to tell it 'acme hebrew and acme tai and acme new fixed arabic' are all bits of the acme font family, hide all those split names and present them as a whole as acme (using inernal fontconfig substitution logic to select the better part of acme to use depending on locale and codepoint needs).