Google also measure bounce rates and dwell time. If a user follows a link but navigates back to the search engine results page within a few seconds, that's a strong negative relevance signal.

I would argue "not attempting to solve the problem" would be better than that solution? (Thankfully, I really doubt any major search engine actually does that.) As you even seem to get: maximizing clicks is relatively easy and websites often only care about the landing, so there is not just a bad optimization in a search engine doing this but it leads to a horrific incentivize to build more useless webpages people click on and then leave. It is also worth nothing that in addition to that model falling prey to clickbait topping all the search results, it also penalizes results whose snippet contains the answer (which is the best possible search result from the perspective of a search engine being useful to a searcher).

Apple once noted that sorting iOS apps by number and speed of installs was a big problem because it encouraged apps that looked and sounded too good to be true but often were horrible to use, so a while back they switched to a model of "how long do users leave this app installed on their devices" or something like similar (it might have been an "engagement" model, which has the much less bad but still existent well-known flaw of "it encourages people to make their app a little slower and require a few more steps to slightly increase the metrics of how long people use it").

Well, one alternative would be to revert to its old model. This is a direct result of Google's decision to emphasize user behavior over information in its algorithms. As I alluded in another comment, this has created a new kind of link farm.