This is a problem for the early stack. The bug was never triggered before, because I guess there was no trap call from the early stack. In my latest patch for the startup code, I added a Ssystem() call very early on the early stack, so any programs built with the current CVS MiNTLib crashes before main() on systems (like mine) where the bug is visible.
MiKRO, that reminds me the strange problems you experienced when the system is going out of memory. I don't know if it is related or not.
I think there's good chance it is -- I was experiencing that crash exactly as you write -- by calling function for showing error message (i.e. trap #1).