While that shows a good use case for a minority, it doesn't refute at all “zero benefit to the overwhelming majority of end users”. Most end users don't have configurable keyboards, and even less program microcontrollers.
This is why it is contentious: On the one hand you have a great QoL benefits for a minority of users who are more than willing to accept the potential fingerprinting risk and consider other dangers to be overblown hypotheticals, or are technical enough to mitigate those issues. On the other you have the vast majority who have no need of the feature at all, and are probably unaware of any risk and will be until it becomes apparent in a can't-be-undone manner. Google sides with the former, Firefox and Apple err the other way. At what point a small potential risk for everybody is worth QoL benefits for a relative few, is the main point of contention (the second largest probably being a mix of “is the risk really a risk anyway?” or “but you'll be fingerprinted to buggery & back anyway so what difference does this bit-or-few of data really make?”).
A side issue is the concern about the browser becoming a bloated almost-but-not-quite full OS, and a huge mess that needs much effort to maintain to keep its growing attack surface defended.
Besides making flashing esp32 devices easy there is also another niche use-case for webusb and that is Web MiniDisc; keeping those recorders alive long past manufacturing EOL.
This is why it is contentious: On the one hand you have a great QoL benefits for a minority of users who are more than willing to accept the potential fingerprinting risk and consider other dangers to be overblown hypotheticals, or are technical enough to mitigate those issues. On the other you have the vast majority who have no need of the feature at all, and are probably unaware of any risk and will be until it becomes apparent in a can't-be-undone manner. Google sides with the former, Firefox and Apple err the other way. At what point a small potential risk for everybody is worth QoL benefits for a relative few, is the main point of contention (the second largest probably being a mix of “is the risk really a risk anyway?” or “but you'll be fingerprinted to buggery & back anyway so what difference does this bit-or-few of data really make?”).
A side issue is the concern about the browser becoming a bloated almost-but-not-quite full OS, and a huge mess that needs much effort to maintain to keep its growing attack surface defended.