There are few questions in here, so lets take them in the order you have them and see if we can straighten them out.
Right. The G4EB2 does not have the EB1 counting chip, so it cant do any on brain counting.
Right. The lack of the counting chip does not change the brain scan rate. In the case of the EB1 think of the digital input modules going directly into the counting chip, so the counting function has nothing to do with the scan speed or TCP/IP load on the brain. Regardless, the scan speed for the event reactions etc will be the same for an EB2 as an EB1. Its just that at any stage your strategy can ask for the count and will get an accurate number from an EB1.
Cant say about 'emulating' a pamux system with an R1.... That's a loaded question. Pamux is really fast. Is the 486 running anything other than DOS? How busy is the network? Remember, every single TCP/IP packet on the network is going to demand CPU cycles from the controller / brain CPU to check it for its MAC address. Thats why we put the '0xFFFF F038 0294 | Digital Feature Scan Interval (msec) | 1' setting in. It used to be possible to 'swamp' the TCP/IP stack and since that has a higher priority than the digital or analog scanner, the controller/brain would not poll the point data often enough and you could end up with stale data... So, perhaps now you can see why we cant straight up answer your question on this one..... Pamux did not have things like TCP/IP to deal with.
Cheap, fast, powerful. Pick any two.
Some good questions that got into the nitty gritty of the controller / brain there Barrett. Sure sounds like an interesting application.
Edge case applications are the most risky, and the most rewarding.