I once was a bit too idle and created a program which died with all memory and registers zeroed and all three possible errors. The only flaw was the PC being after the end of memory rather than 0. That was on the 6400 which had almost no pipelining.
The 6600 had a bug which the 7600 might have had too. Instructions were 15 or 30 bits, with 6-bit opcode and three 3-bit register numbers. The conditional instructions used one of those register fields as the condition number (zero, non-zero, negative, plus, nan, etc) and if that happened to be the same number as a busy register, the instruction would stall until that register was done, even if not involved in the instruction itself.