I got to say, to me it seems rather ... unlikely that this is used on a wide scale. The differences in mass produced hardware must fall within a fairly small distribution, and they only mention rise-time as a measurable variable. Taking into account other variations like temperature and device aging, and I have a hard time believing that these fingerprints are highly unique or stable over time.
In addition it seems unlikely that your average receiver hardware could even pick up such details. These are super low-level details observable only at the lowest level of the radio stack.
There's also no sources cited that this is used on a wide scale. To me, I would put this in the bucket of hypotheticals (like the power company stealing your data from your PC based on voltage fluctuations from your laptop). Perhaps barely physically possible, but not really worthwhile.
(unless of course the fingerprints are put there intentionally and secretly... like the yellow dots from printers)
I know nothing of radio communicaion, but having slight variations on some part of network negotiation might limit conflict between devices. The same ways you would add some radomization on broadcast responses on networking, or auto-reconnect, but in a low level, on hardware manufacturing. This could be a easy to track, recognizable signature with standard antenna hardware.
I feel like this is the TV detector van[1] equivalent of cell phone companies. It might be possible to implement, but I doubt it's widely implemented as the article suggests, if at all.
[0] https://en.wikipedia.org/wiki/Radio_fingerprinting