Membrane keyboards are different (different type of contact and the actual contact action is slower) so are not affected by switch contact bounce to the same degree.
Also the problem of contact bounce appears to get worse over time, or with lack of use (but this is opinion I don't have any evidence to back this up). Part of this may well be contamination and dirt on the switch contacts.
With keyboards, the software (firmware) will have a switch contact de-bounce routine. But of course the firmware for the Zeddy was written for the membrane keyboard, and when Memotech made their keyboards, the switches were nice and new. So there was either no problems, or keys repeating were less frequent than now maybe? Not having a Memotech keyboard back then I have to ask for the help here, as I can only guess

It should be possible to write a better keyboard scanning routine that can compensate for the contact bounce, that should work with most (if not all) keyboards.
Now buffers:
The term buffer has two completely different meanings when used with electronics and computing.
1st - to act as a temporary store for queued data (what software "people" first think of),
2nd - to isolate, amplify or boost a signal, for example the transistor used in composite video modifications is a buffer (what electronic hardware "people" first think of).
When Memotech describe their keyboard interface as a keyboard buffer, it is the 2nd of the definitions that they mean, because inside there is a logic chip that isolates and boosts the signals between the Zeddy board and the external keyboard.
Mark