Discussion about the Korg PolySix synthesizer

This has come up in various degrees with several boards so I guess this
could be a more common problem.

Short description: MG Delay doesn't always work reliable

Long description: I'm getting different times for the MG Delay depending on
the time between key presses. If there is a longer time between key release
and new key press (several seconds) the MG Delay time will be significantly
shorter.
It should be noted that this is also present with original KLM367 boards
albeit to a lesser degree. For example, one original board I have has a
"regular" maximum delay of about 9 seconds; this drops to about 7 seconds
under the above scenario.
However with the clone boards the behaviour can be more severe. I'm having
one that behaves about the same as the original board; however another one
has a maximum delay time of about 7 seconds which drops to less than one
second in the above scenario.

I suspect T5 (Q5) to be the culprit here; swapping this with another one
resulted in a maximum time of about 7 seconds and a minimum time of about 3
seconds. However swapping this with the original 2SC945 (of course I took
care of the correct orientation) resulted in near perfect behaviour.

So the question is if the 2N3904 is really a good substitute for the
2SC945? There are of course some tolerances; however that those tolerances
result in such different results strikes me as odd. Or is that batch of
2N3904 really that bad? What have the other clone builders experienced?