> ...  The obvious solution is to have a
> variable somewhere that always mirrors the last value written to each
> write register (presumably an array of chars, one per register), but
> the driver doesn't seem to do that now.  Am I missing something?

Nope - you're not missing anything.  But there is some work being done on 
the serial driver that _may_ provide exactly what you're looking for - a 
"copy" of the last value written to those darned write-only registers, as 
well as values that need to be written the next time it is "safe" to do 
so.  Just don't hold your breath - it's currently just alpha (or even 
pre-alpha) code right now.