This is a very good counterexample to the article. USB is starting to become "the one" physical interconnect for consumer hardware of any kind, and I believe that is a great thing. Having the ability to connect an eGPU docking station to the same port you normally use for headphones while on the go increases choice a lot.
Specialized standards cannot ever be a one-size-fits-all kind of deal, but in the scope of generalized standards like USB, which abstracts serial interfaces to a point where you can literally use them for anything, convergence to a single one needs to happen. The thing is that most standards are not part of that generalized group. And those that are wouldn't even come to mind when talking about standards at this point, like the x64 architecture.
Then we also have stuff like ATX12V which is the one standard for internal power delivery in PCs and will be for a long time to come, but that also serve as an example that things could be done much better.
Doing non-standard things is important, though. I think in many cases it would be useful if non-standard interfaces/procedures were never standardised, but just well-documented so the best one will become a de-facto standard at some point. The NCASE M1 reservoir measurements have good potential to become a standard for SFF reservoirs, for example, and people (including myself) pushed hard for it to be supported by the Cerberus.
