Implementation details at layer 1 and 2 of the OSI model. If machine aa needs to talk to machine zz (and there are all the machine in between them), then aa needs a timeslice to change the electrical levels that are on the actual ethernet cable, as does zz in order to reply, but if ab and ac and ad and all the way to zy are also busy communicating, then when does zz get that timeslice? Switching (vs hubs) improves the situation drastically, but as you scale up (like, thousands of thousands aka "webscale", there's clearly going to be some sort of limit at some point (until you apply other technology). What is that point using Ethernet though, and how do you calculate that?

The OSI model never really panned out in reality

Except it did, and that is the point of Layer 3. Also the esoteric nature of OSI doesn't mean it isn't useful.

What I mean is that the OSI layered model never got implemented exactly to spec.

Yes good point