A socket is a dedicated tube for communications back and forth between server and client (the website) all within a single connection. A socket is nice because it’s an easy interface for sending arbitrary data either direction without the client needing to know to explicitly request information from the server. That makes it easier to have a application which responds to changes rapidly. The way Http worked traditionally didn’t allow for servers to push data because Http has a single request and response per connection and clients usually have no ports open (required to receive connections), meaning clients need to request data.

Websockets are a technology built on top of later versions of http and allow a “fake” socket interface which allows servers to push data without the client explicitly requesting it as long as the webpage is open. In this case the server was either sending back the wrong information to the wrong users or the client got mixed up about what to do with the data the server was sending. Either way the bugs lemmy was seeing isn’t intrinsic to Websockets and was purely an implementation issue. The downside for Websockets is it can be a lot of information and heavy overhead to support if not implemented carefully, in lemmy’s case the developers decided to give up on it rather than spend time trying to fix it due to the sudden increase in users and demand on servers. It could potentially return in the future, hopefully in a more polished state.