Having heard back from Rigol (via our good friend Drieg), it appears I was mistaken about the following bug:
When both channels are on, any memory read that requires more than one packet of data will ALWAYS fail. The maximum packet size is around 2MB - so any memory depth >2MB for byte mode or >~200kB for ASCII mode can not be read correctly.In fact, it works fine. The thing that was confusing my software is that, with a single channel on, the two SCPI commands used to test and then read the buffer (:WAV:STATus? -> WAV:DATA?) both return the same number of bytes to be read from the DSO. But with two channels on, the commands return different numbers, with the WAV:DATA (not STATus) command containing the correct number (i.e. the DSO has filled it's buffer further after the STATus command).
So I've removed this bug from
the bug list - which, BTW, is reportedly down to virtual NIL in the upcoming release.