Apologies. My fault entirely. With my current physical condition (I broke my right wrist Monday) writing these tech tips takes forever one-handed and is physically challenging. My post about system resources (in different posting) took over 30 minutes to type.
In order for a system to run at a certain speed all components must be capable of that speed. To simplify - slowing down the processor allows all components time to react in the proper manner. RAM, for instance. Let's go with the classic design, not taking into account current clock double and quadruple-pumping designs.
The RAM can only accept data in an exactly timed window between the down side of the clock pulse and any other timings, like the refresh and the address strobe. Reaction and storage (or retrieval) must occur in this window. If the RAM is off-timing it will not catch the full rise, or fall, of the data square waves. So, for instance, instead of seeing the pulse from the beginning - rising from 0 to say 4.7V which would read as '1' - by being late on sync it would see the pulse falling from 1.8V to 0 and read it as a '0'. Instant bit error.
Sorry for the technical, but the only way to properly explain. Previous based on 5V dynamic RAM, single clock read.
So slowing the system down allows all components to see and react properly to all events. Since he stated in the opening remarks that slowing the system down makes it work the direct correlation is that some component can't handle the speed - it either wasn't designed for it, was designed but out of spec (from factory or aging), or suffered damage and is no longer capable of the speed. Your mileage may vary...