Although I don't know of a way to solve this for videoconferencing, historically one way to mitigate the radio/vsat "batchiness" issue and its effect on end-to-end latency was to use a caching proxy server connected to a, er, "real" network somewhere, preferably as near as possible to the headend/uplink station. The modern web's move to TLS also means this technique is becoming pointless even for HTTP/S (although MITMing remains a way around that - many a HOWTO abounds, describing how to do this with Squid).
FWIW, some very old radio systems behaved similarly... albeit not with 600+msec latency :-/. Some of the really old asymmetric TV systems (dial-up for uplink, CATV for downlink) exhibited similar characteristics and were similarly
difficult to mitigate.
Good luck!
Adam Thompson
Consultant, Infrastructure Services
100 - 135 Innovation Drive
Winnipeg, MB, R3T 6A8
(204) 977-6824 or 1-800-430-6404 (MB only)
athompson@merlin.mb.ca
www.merlin.mb.ca