> P.S.  In my thinking about this problem I stumbled upon the following:
>
> delay64 call    delay32
> delay32 call    delay16
> delay16 call    delay8
> delay8  call    delay4
> delay4  return

For reasonably short delays, I prefer to use: [though normally I only expand
it to Delay1024]  Length is 36 bytes, but it uses no F-registers and calling
it even at the 128K level is usually manageable on the stack (6 locations).
When called for more typical delay values (e.g. 2048) the stack usage is often
well within what's available.

Delay131072::   call    Delay16384
DelayXXX::      call    Delay16384
DelayXXY::      call    Delay16384
DelayXYX::      call    Delay16384
Delay65536::    call    Delay16384
Delay49152::    call    Delay16384
Delay32768::    call    Delay16384
Delay16384::    call    Delay2048
Delay14336::    call    Delay2048
Delay12288::    call    Delay2048
Delay10240::    call    Delay2048
Delay8192::     call    Delay2048
Delay6144::     call    Delay2048
Delay4096::     call    Delay2048
Delay2048::     call    Delay256
Delay1792::     call    Delay256
Delay1536::     call    Delay256
Delay1280::     call    Delay256
Delay1024::     call    Delay256
Delay768::      call    Delay256
Delay512::      call    Delay256
Delay256::      call    Delay32
Delay224::      call    Delay32
Delay192::      call    Delay32
Delay160::      call    Delay32
Delay128::      call    Delay32
Delay96::       call    Delay32
Delay64::       call    Delay32
Delay48::       call    Delay32
Delay32::       call    Delay4
Delay28::       call    Delay4
Delay24::       call    Delay4
Delay20::       call    Delay4
Delay16::       call    Delay4
Delay12::       call    Delay4
Delay8::        call    Delay4
Delay4::        return