However, if the optimization for apps is based on usage, reviews, downloads, etc, then having a good experience is consistent with having good rankings.
Not so much with SEO where things like engagement and usage data don't factor much into rankings since it's not reliable
that's a story the link-counting-tool providers (i.e. seomoz) and linknetworks / linkbuilidng companies have been telling for years. and it is bullshit to keep the (crapy) status quo in SEO (because their business is deeply rooted into the status quo).
ux metrics (serp clickthrough, long clicks) are an a factor for ages, see i.e. the discussion about "long-clicks" in the exgoogler book "i'm feeling lucky".
plus: on big sites it is very easy (and i and other have done it multiple time) to test this UX factor. just make the UX of google referred visits significant poorer (i.e. via cloaking or performance delays) and see what happens to your overall SERP impressions and click throughs.
the same goes for reviews, there is even an publicly available googler paper about it.
Matt Cutts and the Google team themselves have publicly said that they tried looking at user data for rankings and determined it just wasn't a reliable measure. Think about stuff like finding addresses.. it's GOOD if a visitor spends 5 seconds there vs. 5 minutes because it means they found what they wanted. Sometimes usage data is misleading.
just test it (it's very easy to test on sites that get some decent traffic). well, "it's very easy to test" if you are willing to loose some traffic for some time (and as a SEO you MUST be willing to so such tests)
also i know the webmaster videos as well, they are more in the tone of "you have to be careful when you look at user behavior".
also your argument has huge holes: "it's GOOD if a visitor spends 5 seconds there vs. 5 minutes because it means they found what they wanted." well, there is an statistical technic called "standard deviation", if the SD is too big, don't factor it in. (i don't say that google does this, i'm sure they are more sophisticated, but it's a good counter argument for your example)
"Sometimes usage data is misleading." well, then G would just have to determine when it is misleading and with which probability.
i repeat: just test it. (and: don't read to much seomoz.)
However, if the optimization for apps is based on usage, reviews, downloads, etc, then having a good experience is consistent with having good rankings.
Not so much with SEO where things like engagement and usage data don't factor much into rankings since it's not reliable