"Kind of a poor man's stack profiler."

A real In-Circuit Emulator (ICE) would give you a real stack profiler; does anyone know if any of the variants with JTAG and on-chip debug hardware can do it?
(I know that the Triscend E5 can't).

Of course, a stack profiler still only tells you what happened during the test; it still does not guarantee that the test actually reached the maximum possible stack usage...