A manager is doing it by hand. Their POS system has the ability to lock out individual items, and this information is available in real time to the server backing the mobile app (GMA).
The machines that I've seen don't have real-time telemetry, and are not networked.
There was an old effort a long time ago where the machines were on a powerline-networking system (LONWorks) but that is dead and buried. The next-gen effort was ramping up a few years ago and I don't really know where it is now. The predictive analytics company they bought was a big part of it.
Their tech stack is far more bleeding edge than I would have envisioned. I wish they gave an idea of why they chose to use k8s at the edge. I can see how containers are useful for deployment, and the fail-over built into Kubernetes is really nice, but I'm shocked that's the simplest/best way to do it at the edge.
I would have expected some VMWare product, just because it's "enterprisey". Shame on me for judging.
Deployments and monitoring are probably another draw. Same tools server and edge side.
Another project to keep an eye on is AWS's Greengrass which has a bunch of capability overlap, but a bunch of new ideas for edge and fog IoT also. I had a mixed experience with it but at least some ideas bear borrowing.
Not sure if this is the same thing, but I recall reading a case study in 2006 on McD's cloudifying their kitchen equipment. The project was apparently a huge failure and was eventually stopped due to runaway costs.
The machines that I've seen don't have real-time telemetry, and are not networked.