Not sure if I'm understanding you right, but I wasn't saying that my experience is the only one that matters. Just that it's not a thing one notices in practice, at least not under conditions I've experienced (I figure a reader can fill in that last bit for a comment written in the first person). Saying AOSP's is "much, much faster" suggests it would be noticeable and afaik it's not (at human timescales), so I wanted to add that info to the thread
Good point about the USB thing btw. It's obvious to me and the reason why I go one step further and leave USB debugging always enabled now that there's this private key authorisation method anyway (it asks for computers whose key it doesn't yet trust), but indeed a lot of users might follow GrapheneOS' advice without realising
Leaving USB dubbing enabled just exposes a lot of attack surface. And if you use USB debugging you are placing a lot of trust in the computer you are connecting to. You don't need USB debugging to reflash GrapheneOS or to sideload updates from the recovery mode. So, it's not relevant to prevent a device brick.
Good point about the USB thing btw. It's obvious to me and the reason why I go one step further and leave USB debugging always enabled now that there's this private key authorisation method anyway (it asks for computers whose key it doesn't yet trust), but indeed a lot of users might follow GrapheneOS' advice without realising