Thank you srnet and Leif,
Plenty to get on with in those suggestions.
Following further tests on the HC12s, with particular reference to Mauried's suggestions ref time delays, I've identified where extra millis have largely solved the issues.
It would seem that the matrix driver requires extra time to process commands like setting inverse than they do for other less complex tasks.
So, assuming I'm understanding the process involved, extra delay before commencing transmission of the next command after transmitting one of those is required to allow for the slightly longer delay before the resultant ack response has completed, and especially so when more than one Rx is calling back. And that explains why a single Rx unit operated fine but two or more give rise to bottlenecks in the code at those commands, but only at those particularly complex matrix instructions.