Have you tried changing tracks rather than playing through?
Recall the Rossini issue is seen when jumping between DSD tracks, but not when playing from one DSD track to another; I don’t know what happens if you play from a PCM track to a DCD track, I would expect the same loss of the first .84 second.
Recall also playing .dsf via USB only works for tracks after the first; the first is always truncated as when the datastream stops it assumes it is PCM and the delay in detecting and switching to DSD is what is the root cause here.
By comparison, a network connection defaults back to PCM between each track, so it will always truncate the beginning.
The problem (at least on the Rossini) is that it takes time to determine whether a bitstream is PCM or DSD and if DSD switch to the DSD decoder.
However, rather than cache the data that is incoming while the detection and switch occurs as virtually all other DACs do, at least the Rossini either just drops the DSD data received to that point or sends it to the PCM decoder assuming it was PCM.
Imagine if you had to switch flow coming in from a main to one of two pipes based upon whether the content was water or alcohol.
Rather than build in a reservoir to hold a second’s worth of liquid flow while you decided which destination pipe got the flow, you just let whatever was in the pipe spill onto the ground while you figured out which liquid it was and routed it appropriately, or sent one second’s worth of alcohol down the water pipe.
That’s what dCS is doing in the Rossini.
The only solution at the moment is to use playback software that can insert a second of silence before each new track - the equivalent of pouring one additional second’s worth of alcohol into the pipe to make up for the amount you would know get spilled out/misdirected on the receiving end.