I am not relying on wrap around.  I am just throwing the stack away, just
as John Payson says.  Another example:

SendMessage:
        //      code to send a radio message
GetAck:
        // code to receive a new radio message or an ACK to a transmission
        Call CheckTimeout       // this call appears in several locations in thi
s large
code block
        ........                        // more code
        Call CheckTimeout
        ........                        // more code
        Call CheckTimeout
        ........                        // more code
        Call CheckTimeout
        ........                        // more code



CheckTimeout:
        // do we need an ACK to a message?  If so, did we time out?
        If ((need ACK to message) && (elapsed time > X))
                goto Retry              // resend message
        if (jamming signal && jam timeout)
                goto ProcessJam // respond to jamming attempt
        if ((do not need ACK) && (elapsed time > Y)
                goto SleepAgain         // no traffic, go to sleep procedure
        return

Each of the gone to locations is at call level 0: there are no outstanding
returns to be executed.
I don't want tests on exit conditions to the timeout procedure after each
call, I don't have the space.  After all, I have got a 10 channel digital
radio network 'slave' device implemented on a 16C57.

What I want is an emulator that emulates the chip.....  A 16C65 does not
melt down when stack overflow occurs, whereas the emulator might as well.
Fortunately, my 16C65 unit (a loop control unit) has space for me to
include a file containing exit condition tests after every call, so I am
not hurt by this rather severe limitation.  So I'm just nagging.