Only it had happened to me 20 years ago that MAC addresses were recycled. We had purchased over 200 PCs from a local assembler company and we had a few that had network issues that disappeared when we changed the network cards. Cards, that when used on PCs on other networks, worked perfectly. It turned out that in the batch that PC assembler company had purchased, there were a lot of cards with duplicate MAC addresses. You can imagine what happened when two machines appear on the network (and I think it was a hub and not a switch) with the same MAC.
They actually thanked us when we pointed it out, because they were facing this phenomenon with other clients of theirs too.
So even smashing the Ethernet card with a hammer does not guarantee a unique number never to be used again. But it is pretty close.
A better hope (not guarantee) of getting distinct UUIDs is to use the v4 algorithm, from section 4.4 of a document that was published a year after the blog post and a decade after Windows 95. No wonder they didn’t try that.
New network cards? Could it have been a mistake that someone in the supply chain was supposed to program a vendor MAC into the cards but noone did so they had the upstream vendor’s test MAC range?
I remember this with some of the early very cheap Android tablets available direct from China, on many models the manufacturer hadn’t reprogrammed Atheros’ default MAC on the Wi-Fi module (supposed to be replaced with a MAC from either the module or tablet maker’s OUI range), so every tablet had the same MAC.
It sure looks like it now. But back then there was no way of checking this out. The address space was from the chip maker and their name was slapped on the cards too. We were a public sector organization in Greece, who bought assembled PCs from a local box mover, who bought equipment by some other importer, who bought them from … you get the picture. Not really equipped to trace this.
Sounds like a super frustrating experience!