mons[0]

Alex mentioned that loops over mons[] were starting at [0], which
should be [LOW_PM] instead.  I only found two, and the mvitals[] one
was benign.  The special level one might have been too, depending
upon spec_lev's thoroughness--I didn't attempt to check.

Once upon a time there was a possibility of moving 'playermon' from
a separate variable to mons[0], so LOW_PM became the index of the
first valid monster.  Instead, 'playermon' went away altogether.
LOW_PM (and NON_PM) could go away too, but I don't see how reverting
to hardcoded 0 and -1 would be an improvement.  We have enough
problems as it is with "giant ant" turning up in unexpected places
because someone used 0 instead of NON_PM to mean "none of the above".

src/allmain.c

@@ -578,7 +578,7 @@ newgame()
     context.tribute.enabled = TRUE;   /* turn on 3.6 tributes    */
     context.tribute.tributesz = sizeof(struct tribute_info);
 
-    for (i = 0; i < NUMMONS; i++)
+    for (i = LOW_PM; i < NUMMONS; i++)
         mvitals[i].mvflags = mons[i].geno & G_NOCORPSE;
 
     init_objects(); /* must be before u_init() */