🚀 go-pugleaf

Lately I've been unable to execute a certain large (about 600 K text+data+bss)
program on my 1040 ST.  It used to work fine.  The symptoms were as follows:
first when I tried to run it from msh I'd get the error message "fundamental
error", and return to msh.  There'd be less than 1 K free memory left, so I'd
have to reboot.  I could, at that time, still run it bu loading it into the
Mark Williams debugger and typing :e, or from either Gulam or the desktop.
As of yesterday, the debugger trick either generates the message "child
process terminated (0)", or crashes the debugger back to msh (with about 10 K
of free memory).  Running from the desktop gives 2 bombs; Gulam crashes trying
to run it.  
   Again, this behavior is recent.  The program (moria, if you must know) used
to run just fine.  I have remade it from sources and compared the result with a
backup copy (Mark Williams cmp utility): no differences.  I reloaded msh from
the distribution disk.  This behavior occurs with from 0-4 desk accessories
(accessory 1 is the control panel, the others are well behaved too: the idea
was to move the load point around in memory).  I use deskmanager 2 to control
bootup, and also have foldr050.prg and fatspeed.prg (and a clock setter)
resident.
   Does this paint a consistent picture of anything to any of you out there?
If it doesn't give enough information for a diagnosis, where do I look next?
Do I take my ST to the shop?  My hard drive?  Trade the ST in?
   All answers (even character assassination) appreciated.
                                 Steve J.     saj@chinet.chi.il.us 
                                                  or maybe chinet.uucp