I am flattered to hear all that.
I think Said is a lot closer to the '1px is broken' than me, but also that it's not really that important. How many websites have dropcaps at all? They are fun to have, but very optional. And so if you find the maintenance to be a burden, you can simply not do that. And then the core CSS/JS basically never breaks. (All of your problems will come from churn elsewhere, like SSL certificates - a major source of feature creep is the demand that everything be HTTPS now - or OS upgrades or SaaS bitcreep.) So my point is that a simple website is as close to zero-maintenance as makes no difference. It's much more complex than plain text files, but that complexity is widely implemented and is stable.
I too am biased towards desktop and have to force myself to make any use of the mobile version, and check for regressions or friction. I find mobile to be incredibly stultifying for any kind of complex work or thinking. Nevertheless! There are a lot of mobile readers (including many people I respect and would like to keep as readers). So, mobile's gotta be good. At least I think we have managed to create about as realistically good a reading experience on mobile as feasible, and serve as a design showcase there too, to shame everyone else.
That sounds plausible, but I haven't attempted to link visitors across devices. I am not sure if Google Analytics even allows that. (The new one is so confusing and hard to use I've largely stopped looking at traffic statistics.) I think it might be hard to interpret such an interaction, though, because mobile users bounce so fast and clearly so distracted & thoughtless & crippled (eg. because they are killing time on a bus), while desktop readers are much more mindful and 'high-quality time', so they are different in many ways beyond what version of the website they see.
Do very many people actually want to claim to show off the latest web dev gimmick? I'd say not. Almost no one is doing that. Even web devs often have quite plain personal websites or blogs which are not trying to show off their mastery of the latest Nodebuzz.js (best viewed in Chrome 234) feature. When people write about their new Proton static site which uses Netlify to cross-compile from Github pages through AWS Lambdafront functions called on demand with a Kafka load balancer to prerender pages for the wasm.js shim and it only costs $100/month after optimization to host their blog with 50 pages and 50 views/day (aside from that time they accidentally created 10MB of logs and it cost $1,000 but don't worry, AWS was willing to forgive it), I am pretty sure they are not going through that to impress an employer! They are going through that because it's a lot of gizmos and puzzles to play with and 'number go up', and the end-result is irrelevant. They are doing it to watch the gears of the new toy go round, as Norbert Wiener said of the atom bomb makers. Which is fine, as long as you're just getting experience (and also not making a new atomic bomb), but may have side-effects in naive people thinking that any of that is a good idea, much less necessary.
Managers & designers, on the other hand, do need churn to justify their existence. Look at the hilarious trends of companies commissioning new fonts where you can't see the difference with some Helvetica-like even overlaying the fonts, accompanied by thousand-word manifestos rhapsodizing over the design process inspired by the cliffs of California and the sensual humanism of Bauhaus (with a cheeky node to Swiss Modernism, and a dangerously daring serif on the 'g').
Oh, I think I would disagree there. As I was saying, browsers are actually amazingly good at backwards compatibility (perhaps too good). Which makes it easy to decide what browser features to use: if it's Chrome-only, sure, it may be gone in 5 years, but you don't have to be a genius to know "maybe I shouldn't make any major use of a Chrome feature that no other browser wants to support, and I should wait and see". I may have mentioned that we have a rule of thumb: we can use anything with >95% global support on CanIUse.com, and anything else must be backwards compatible/polyfilled or not used at all. We've hit a handful of things where the browsers betrayed us (I'm lookg at you, `<srcset>` and `ping` attribute), which you can find on https://gwern.net/design-graveyard but you know, even mid-1990s `<blink>`, which was never standardized to begin with, worked until 2013! So I'll put it this way: I don't think you can name 3 examples of reasonably widespread features, which were standardized and had >95% CanIUse.com support as of 5 years ago (2020), which are now either broken or rapidly being removed and have fallen <95%. (The only examples I can find are still exotic: https://en.wikipedia.org/wiki/Cache_manifest_in_HTML5 , browser FTP support, and https://en.wikipedia.org/wiki/HTTP/2_Server_Push 2 of them would have been bad ideas to use in new code in almost any circumstance, and I suppose for personal websites you could've in theory used the cache for an offline browseable website... maybe. I doubt many people ever did.)
I think there's some degree of anxiety around perfectionism, but also around the aversion of writing (some people find any kind of writing to be very difficult, even when they have just written an essay by talking to you in a chat client), and the fear of pseudonyms being broken or being canceled. (Zoomers & Alphas seem particularly terrified of 'the dark forest'.)