That is not entirely true. The heat did cause lots of problems but the heat problems were easily fixable.

The biggest flaw with the original design was Williams choice of opto's and using them to drive the switch matrix directly. There is no margin for error when using an opto directly across the Wms switch matrix. The standard opto's had a VCE-sat in excess of 1V **at the high IC-on current used by the switch matrix**. Williams had to have QVE11233.0086 opto's special made and these were just barely under the VCEsat threshold needed for the Williams switch matrix to detect a switch closure. But to get this lower threshold, Williams drove the parts at the absolute maximum limits resulting quite a bit more heat generation and a less than stellar lifespans. Now the special order 0086 optos out there are pretty much used up and are getting quite rare so this makes them difficult to repair existing boards.
Somewhere on RGP there is an end to end analysis of the switch chain using these optos. Can't find it, written too long ago.

RD did the exact same thing that Williams did. RD didn't redesign the board; they simply copied the existing design for the switch matrix - including the biggest design flaw. To get around 0086 opto availability, RD special ordered optos with the same specs.

Proper way to fix this? Look at any recent Stern board that use optos.

The Ingo board is a complete redesign in the switch matrix area. They don't make an attempt to directly drive the switch matrix directly with the opto. They use optos to detect the clock movement but use separate drivers for the switch matrix. Increases reliability tremendously.

All of the new designs replaced the lamps with optos. That is a plus for all of the boards.
I like the LED's used on that Caspers board. But if they put the opto directly across the switch matrix....