So many intangible factors here. A big one being "does improving the overall search experience by providing the most relevant results in the SERP outweigh the opportunity cost of the lost display revenue?"
I wonder how this game will play out though since well-optimized sites tend to be very careful with what they put in the title/metadescription to encourage the click through, and ultimately an entire subset of sites have no incentive to create content if Google starts publishing results that don't necessitate further investigation (ie. clicking through to their site).
YouTube (another Google product) for example shows for videos that contain songs links to iTunes, Amazon, Spotify and Google Play.
Google Search could do the same — list different places to get the content, etc.
It appears as if resolution can only come from splitting up the search part of the company with the rest (assuming it is split into only two entities). I think Google can resolve this by splitting up the knowledge graph results from the rest of search by making the feeling lucky button analogous to showing only what knowledge graph and information cards. They still would have a very dominant share of search, mobile, etc. but it would alleviate the connection of Google the search engine provider and Google the 21st century thesaurus.