I just had a conversation with an iPhone repair guy. He asked me to help him fix a centering problem in his website. So he pulled up the HTML, and it was just awful. Table-based layout, broken tags... But it worked, and I taught him about the <center> tag. He gave me a discount on my repair.
Malformed HTML input is still very much alive and well, and necessary for non-technical people.
Of course, now we've replaced the hundreds of <td> with hundreds of JS files instead. Problem solved!
But it was OK because the forum was running phpBB, which would invariably be hacked/spammed, so you'd be getting a free dialer which was guaranteed increase download speeds ;)
Oh the horror.
It's still often written by hand, but not generally by novices.
Quite the opposite!
Regarding CSS, it rose very gradually, but its rise created a slowly growing problem for Dreamweaver, resulting in the transition from then, when most pros used it for at least part of their workflow, to now, when most don't. Dreamweaver had the best WYSIWYG mode of any HTML editor at the time, and it had a pure code editor, and its big claim to fame was that you could edit in either, jump back and forth between modes, and they would (mostly) remain in sync. You'd do the rough layout visually, then tweak the details in the resulting code, then tweak a bit more visually, then add another touch of code, etc. As non-HTML components on the front end gradually became more important, people used the WYSIWYG mode less and less, gradually shifting toward straight coding, significantly reducing the value of DW's dual-mode advantage over pure code editors.
Oh but there was one, and it was desperately needed!
Gruber's original implementation was absolutely riddled with logical inconsistencies, outright bugs, and even an escaping vulnerability that was exploited by an XSS worm to self-replicate on Reddit.
This project originally started as a PHP transliteration of Gruber's pile of regexes, and it's early version history is a nice summary of all the warts as they were discovered (and never fixed upstream): http://michelf.ca/projects/php-markdown/classic/#version-his...
To elaborate: In the very early days, HTML was often written by hand. Part of the early vision of the web was that anybody could write HTML and publish content.
It was thought that allowing malformed HTML to result in a completely blank page or an error message would be too unfriendly for less technical users.
In those early pre-CSS pre-JavaScript days, HTML was more like bbcode, Markdown, or RST today. It's interesting that AFAIK those languages don't have a notion of malformed input -- there is no movement to tighten the bolts on Markdown in the way that XHTML attempted to tighten the bolts on HTML.