I 100% agree with this, and if you look at the thread you’ll see a good portion of it is making fun of @HRose for exactly that reason.
That said, though, blaming the user only goes so far, and that thread is just one example of sustained discussion on a topic not explicitly preordained at the thread’s genesis.
Here’s a different example where you can’t get hung up on the user error. I started a thread about the topic of digital CCGs:
Some ways into that thread, someone mentioned a new game in the genre, War of Omens, and we talked about it for a while:
But if a new player just picking up War of Omens was curious if people had talked about it, again, the most relevant result is buried by a bunch of marginal search matches. Discourse thinks that a post in a big thread that contains ‘war’, ‘of’, and ‘omen’ separately is a more relevant result than four verbatim matches of the whole search string in the span of nine posts in my CCG thread.
The answer there can’t be that it’s incumbent on the thread starter to go back and edit in a mention of every major discussion topic that gets broached.
(And yes, I know that this example isn’t perfect either because you can put the search string in quotes to force an exact match. But even if each of those threads did contain exact matches, Discourse would prioritize a single “I’m going to try out War of Omens this weekend” over a dozen posts of meaty back-and-forth discussion about the merits of the game, just because the former occurred more recently and in a larger thread.)
I know that search isn’t an easy problem, but we’re not talking about building a deep neural net to learn a better PageRank, here. Just manipulating the comparative weights placed on simple, human-understandable concepts such that some number of exact-search-term verbatim matches have a chance to outweigh an isolated mention in a large and recently updated thread would be a huge improvement.