* Fail a project in Java, and you have a project failure. Fail a project in Haskell, and you have a Haskell failure.
* Management has a large veto power, and management usually favors technologies they are used to, or used to use themselves.
* Management also favors their ease of work with the tech. In practice, they often hold the belief that it's much easier to find java developers than it is to find haskell developers.
* Haskell suffers from a reputation of being "academic" and "hard to do real things with".
In practice, I've done great commercial projects in Haskell, and the technology was worth using. When done right, recruiting for haskell projects was much faster than recruiting for any mainstream language. On the flipside, it was sometimes hard to get adequate corporate stewardship for the language (for instance, managers/HR insisting they found no one for the job after publishing through the usual channels when you know half a dozen people at the local meetup that would jump ship if offered a position).
Other developers working in related areas are using Ruby and Python but my software runs a lot faster.
This is mostly because the majority of software developers only know Java and are averse to learning new languages. Also, companies chose Java at one point and are now taking next to no risks by introducing new languages (not even languages on the JVM). They can also just fire and rehire Java developers because there is a large pool of available developers that they can just plug-n-play, so to speak. They're not interested training people.
Part of that may be the continuing influence of PG: I both learnt Lisp and found HN because of his essays, and I suspect I'm not the only one.
But also, the HN community generally has an academic fondness for new and niche languages, so a language's popularity on HN is a poor proxy for its popularity in the business world.
(A last point: I think the diminished importance of Lisp in the last decades is less due to an inherent flaw of the language and more to what you might call "accidents of history" - particularly its close association with early attempts at AI.)