For the font handling within the wscons framework, I'd propose
the following:

-Change the "load_font" member of "struct wsdisplay_accessops"
 to take a string only as argument. This string identifies
 a font to _use_ for a certain screen. There wouldn't be a notion
 of "loading" a font into a specific adapter anymore. Adapters
 have to use builtin fonts, or to get it from "wsfont" or another
 (private) source of font data.
-Extend "wsfont" to take over the "load font" function which was
 per-adapter. Create a "pseudo-device" for the cdev interface.

Afaict, only the "vga" driver uses the "load_font" entrypoint
at the moment. The change would allow to integrate "vga" better
with the "wsfont" framework and to make things cleaner in general.

What I've tested so far is the first part; I'm running a "vga"
driver which takes its fonts from the "wsfont" pool. If
the needed fonts are compiled into the kernel (options FONT_xxx,
see dev/wsfont/files.wsfont), screen types like 80x50 and 80x40
can be used without loading a font from userspace before.

Does anybody see problems with this direction of development?

best regards
Matthias