I created an upstream issue in Chrome for this: https://bugs.chromium.org/p/chromium/issues/detail?id=849106

Adding an example as what it looks like on iOS Safari after discussing with @Jhernandez
https://webpagetest.org/result/180607_7X_89696060282ad9321ec2a3846b893d33/1/details/#waterfall_view_step1

Clearing out old Blocked column.

It seems Chrome might have improved to have a more optimal approach like the one we see in Safari.

But, I'm not sure. I tried quite a few different article sizes (to vary HTML payload size), and different network emulations (to vary artificial latencies) and for most combinations I tried we still see the waterfall the same way as Peter first found it.

I'm thinking that maybe with today's Chrome, it still often looks like the CSS first byte (and visual paint) are delayed by HTML completion, but maybe that is a coincidence? If the HTML is short enough, it will complete before the render can happen. And if the artificial latency is generally long, then it will take longer for the CSS bytes to arrive and thus arrive shortly after the HTML finished.

On a large article (like Obama), with a medium speed, I found that we seem to get the "ideal" where CSS and paint both finish completely while the HTML is still being downloaded:

I think this is still an issue. To test and verify I pushed some tests to run continuously and forced them to use HTTP1, I only did it for emulated mobile, and the test use simulated 4g. Make sure the test type is http1.

https://grafana.wikimedia.org/d/IvAfnmLMk/page-drilldown?orgId=1&var-base=sitespeed_io&var-path=emulatedMobile&var-testtype=http1&var-group=en_m_wikipedia_org&var-page=_wiki_Barack_Obama&var-browser=chrome&var-connectivity=4g&var-function=median&var-s3path=https:%2F%2Fsynthetic-tests-result-wikimedia.s3.amazonaws.com

Then you can compare with the same URLs and change the test type to firstView. The impact is bigger than I thought on not so slow connection.

Here's two HAR files, one with HTTP1 and one with HTTP2, you can drag and drop them both into https://compare.sitespeed.io and then use the slider to see the diff.