Are you talking of the very poor supply rail decoupling, the use of 16 bit ISA when it looks like it was only an 8 bit use, or the QFP devices that were pretty much guaranteed to be running hot and working their way out of the holders.
The 'decoupling.' Those "little blue caps" are 0.1uF polyester. *Someone* didn't understand the differences between polyester and ceramic caps, for decoupling. He also didn't know about not feeding TTL outputs into 74HC inputs, and most disastrously, had no idea that one shouldn't pass unbuffered logic outputs (and the logic ground) straight to unshielded cables going off board, to other devices in the upright pokie cabinet. Which players often like to jigger with piezo spark lighters, and whatever.
There were other problems too, but I'd have to dig out my notes. If I even still have them. Oh, one was, the CMOS battery-backed RAM (which holds critical data) wasn't adequately protected from unexpected power downs.
The board went into production and was hopelessly unreliable in the field. This happened because no one did any peer design review, management had no idea, the machine testers didn't bother to act like bastard players, and there weren't any lift motors etc near the testing area.
Redesigning a stupid ISA-based IO board, because someone else did a crappy job, wasn't much fun. Pic of replacement board. 4-layer, but most importantly fully isolated I/O signals and ground. This one was reliable.