Regrading modules: if you want to get these modules fast, you pretty much have to buy from Amazon with prime shipping. These modules are also available from Aliexpress.com for much cheaper price but those ship from China and can take weeks. I am only aware of one type of W5500 module:
https://www.amazon.com/ARCELI-Ethernet-Network-Hardware-Microcontroller/dp/B07JLFN3T1
On the other hand, ENC28J60 has several variants, but only the following two have 2×5 pins that match OS 3.0 design:
a wider module: https://www.amazon.com/ENC28J60-Ethernet-Network-Module-Arduino/dp/B01FDD3YYW
a thinner module: https://www.amazon.com/ENC28J60-Network-Module-Schematic-Arduino/dp/B07C2QNGCC
I still think the issue with ENC28J60 can be fixed in software. For one, we know that all OS 2.x used ENC28j60, albeit with the EtherCard library, and I don’t think lockup is a common issue I am aware of with OS 2.x. So it probably has to do with UIPEthernet library. Also, at the minimum, if I can find a reliable condition to check when lockup has happened, then I can have the firmware trigger a software reboot, and this can be done in a program-safe way (i.e. it only reboots when there is no program running). As long as this doesn’t happen frequently, it should be a reasonable solution. At the moment, though, such ‘condition’ is very elusive, because as I said, when the lockup happens on my test unit, the microcontroller still runs, time is correct, programs run, link status is fine, buttons still work. So the condition to flag lockup would have to be from reading ENC28J60’s register values to figure out a consistent pattern. Another way which I will try is to periodically issue a ping from the controller to router, and see if the ping times out.