useHighTrafficDrive logic in K* classes code.

Hello,
My question is about Wisconsin and K* classes code.
Did anybody observe the notably by eye short delays of avatar or space objects movements in KAT demo (more than one participant network/single machine mode)?

So, from the first view, the problem was lying in new useHighTrafficDrive logic (KAvatarUser). So, that, if we switch it to true the delays disappeared.
KAvatarUser>>
#useHighTrafficDrive
    ^HighTrafficDrive isNil "notNil - by default"

but, the new problem was occured, and it is strange(?):

Example scenario (all participants has high-end machines, so no lack of performance):

[1 step] Run at first machine KAT demo
after the delay[Waiting for connection..] of ~10 sec ( in comparison: less then 1 sec when: 'useHighTrafficDrive = false')
avatar is adeded to active space (CroquetHarness #renderWorld).

[2 step] Run at second machine KAT demo
the one way to break the endless delay [Waiting for connection..] - is to (?)refresh the SqueakVM OS window(resize or minimise).
after that avatar is adeded to active space.
and then everything is ok (all movements are smooth on all network participants).

I've reported it as a bug ( http://bugs.squeak.org/view.php?id=6539)

Thanks,
Nikolay Suslov