They are only a failure if adoption was a goal. I am being serious here. It takes significant time and money to get language adoption. Every major language, C++, Java, C#, Python has had significant financial and resource backing.
Often the financial backing is very conservative. It does not like to take risks. So they go with what is familiar.
Second, I can't think of a single academic I know that set out to create "the perfect language". Often one is created to explore an idea. That does not mean the language isn't incredibly valuable for the lessons it provides - but it likely was never meant to hit mass adoption.
> They are only a failure if adoption was a goal... Often one is created to explore an idea.
Hence "academic". Design for idea exploration rather than widespread adoption is a crucial property of a language, and a valid target for criticism.
Some PLs, are therefore academic in the same sense that Esperanto is academic, and are rightly criticized for that. They may be "perfect languages", but the fact they are not widely adopted is conclusive proof that they're a failure, at least in some respects. Sometimes this failure could be mere marketing; usually it's much more than that.