I worked at Yelp for five years and was there when Yelp engineering first started doing "hackathons". People worked on some interesting stuff during them, but very little of it was meant to be production ready and useful (vs toy) projects saw little exposure outside of the company or the engineering department itself. Lots of things were meant only to scratch our own developer itches or impress co-workers, like internal engineering tools (which have their own value). I attribute this to "yelp the public facing product/service" having a master plan that is determined outside of engineering proper—nothing wrong with that, it is what it is. So if you had some cool idea to, for example, revamp the UI/UX of the business photos page, you could work on that during the hackathon, but it was only a hackathon project. This encouraged people to work on throwaway toy projects or experiment with deep backend stuff during hackathons to blow off steam vs something that someone could point to publicly. Not to say that no hackathon projects ever were exposed non-internally, but it was a very small percentage. Some people used hackathons to catch up on regular work while many other people were working on hackathon projects. Maybe it's changed since I left.

"Hackathons" are not a silver bullet solution to trying to get people to have ownership over the solutions to problems they are solving for themselves, their department, and the company.