If they are on different sites, each with their own firewall/router, then yes, I can see why you would be looking at using MQTT on each R2. That’s the perfect use case for pushing (publishing) data from each site back to a central broker.
I know of at least one person that wrote their own MQTT chart for the PAC controllers using comm handles. They did not do any SSL security nor did they implement SparkPlug, but it was MQTT from a PAC none the less.
If the PAC R2 Controllers are on different sites, then you for sure do NOT want to do a poll architecture and start opening ports, don’t even think about it.
Other options for pushing data would be streaming. You could publish MMP data from the PAC R2s to a central PAC.
We have a lot of customers use this approach to push tables of data from remote sites to a central PAC S controller. Each remote site has a block of MMP scrachpad area mapped out in the PAC S.
Of course that requires a fixed IP or host name at the central location and a port forward rule to the PAC S, so that’s less than ideal as well.
Have you gone over the Opto 22 SNAP networking guide?