No experience, pure conjecture based on my semi-informed impression about the capability of cirquitry available over the last 30 years.
My impression of the Sony Oxford was that it was built as one massively powerful DSP core along construction principles found in supercomputers like the Cray at the time, while the (related) DMX-R100 was built with a multiple of RISC or SHARC processors that worked on specialized tasks and separated datastreams. The Oxford could route anything anywhere, and it was all done in software, while the R100 had some limitations because the flow of datastreams were to some extent hardwired. These two desks represented not only two extremes of the pricing scale, but also two different directions of hardware and how that influenced flexibility.
I don't really know anything about the internals of the X32, and how many DSPs share the work, but am assuming there is a central DSP that communicates with an IO section and that there might be a software controlled multiplexer or matrix between the two. Routing, whether that is rearranging chunks in a serial stream or connecting a n-wide bus through a matrix is in all probability done by setting bits in a n by n table.
I can understand that at some point, the streams from the ADs are likely to be routed by the stream size, ie. if the AD is a 192KHz device that will do 4 channels at 48 KHZ, than the routing will be done by 4 up to a point where the stream will be split or multiplexed further, (or by 8 if it is a 384KHz device), and that if indeed the X32 is based on processing of several 384KHz datastreams, a lot of the routing will be defined by the streams, and thus the limitation will be due to hardware architecture.
Anyway, I'm just rambling, pay no attention
~
~:smile: