AM5728: Unorthodox uses of PRU and Ethernet PHY

The Harvard Physics community is doing a Big Rethink about how they control their benchtop and room-sized experiments, to simplify and consolidate all of the custom communications protocols in use.  One typical request is to be able to send a 16-bit word across a lab (50 feet, say) through an isolator, and get it back again with a round-trip time of about 1 usec or so.  In other words, they need to send much less data with much faster latency than, say, UDP.   They typically solve this with various SERDES chips, SONET hardware, etc., and want me to come up with a "standard" solution.  It occurs to me that connecting two Sitaras, with PRUs bit-banging into Ethernet PHYs some very simple protocol ("Here's a word.  Now send me one."), that could just about do it, because 100baseT sends 25 M nybbles per sec between the Phy's, bidirectionally, and isolated.  So before I reinvent the wheel and commit a large research university to yet another custom "standard", I wonder if there's something out there close to what I'm looking for.  Any thoughts?