Is there any actual evidence that might be true? I encountered this argument many times, but only from programmers (who thought it was to difficult because of their tech stack), never from users.
As a user, there is no site that does not have search, no matter how good it is, that does not feel quite horrible because of that. Even the search by Google feel horrible -- romhacking comes to mind. It is also frustrating because some instances of title searches are notoriously unindexable, while a nice site search lets you narrow the context through tags or model specific attributes (system, genre, ...).
The statement is intentionally hedging and qualitative. Some people do the usage pattern, some don't.
The advocacy is to push back on search as a difficult and often unnecessary problem to solve.
The caveat is my increasingly toxic pattern to push back on almost everything as unnecessary. It's probably overly antagonistic.
I'm a techno pessimist programmer. I didn't understand this attitude when I was younger but then again people often become that which they fail most to understand
Formulated like that, I'm very sympathetic to your point of view. Having a small set of well-refined features rather than a growing spaghetti of half done ones.
I guess my own point of contention is that programmers would say "task X is hard" when it isn't that much, it is that their technical choices make it hard.
For instance, one Front-End engineer was really proud of using a very recent and trendy framework to rewrite the whole site from scratch. After deployment, people are ordering the wrong products. His reply was, "keeping the query filters in sync is too hard of a problem". It's really not.
As a user, there is no site that does not have search, no matter how good it is, that does not feel quite horrible because of that. Even the search by Google feel horrible -- romhacking comes to mind. It is also frustrating because some instances of title searches are notoriously unindexable, while a nice site search lets you narrow the context through tags or model specific attributes (system, genre, ...).