It's very encouraging to see documentation of failed approaches to problems in technical write-ups, even if brief — for example, in How to speed up the Rust compiler in 2022.
Anecdotally, at least half of the big interface/architectural design ideas I'm trying these days (and many more small approaches to performance debugging, etc.) don't end up working out, and the important ones always get a postmortem in my notes for future reference. They've proven invaluable when discussing these issues with other people or when attempting later solutions, because I can look beyond my imperfect memory to see what issues I must design against.