First of all, I played with some web frameworks in Haskell and didn't like them at all: the yesod community had a very dismissive tone towards people that didn't want to use what they perceived as "standard" and Snap is well organized, but sometimes weirdly implemented around the corners (WAT? The templating library also serves static files?). I would love to use any of them, but they are still not fit to build ecosystems like Django, Rails or Spring.
Its nice that you picked the JSON library (by Brian?) as an example for a lib that can be written in "a few day". `aeson` is a great library. Its just not something that someone without detailed Haskell knowledge can write in a few days. It requires good knowledge of Haskell, especially things like the GHC.Generics library and Template Haskell (Hint: GHC.Generics is not very well documented for practical uses as well, the documentation spares you of the hard cases).
I found that if you go away from those 'core' libraries, quality drops sharply. Giving just one example may be unfair, but I do it anyways: the readline bindings (haskeline) didn't build on OS X for 2 major releases straight. Also 0.1 and 0.2 seem to be very common version numbers on hackage and most documentation boils down to type descriptions. For loves sake, I couldn't find a good and proper tutorial working with bidirectional streams (think websockets) and asking a GHC committer ended in "well, we're still not sure what the best abstraction for this is, theres multiple competing" and someone on the Haskell User Group Berlin said "well, its not such an easy problem, but it will be solved in a year or so". In Ruby, thats implemented in an hour (safe and sound using libraries, with proper abstractions), today. Yes, I honestly think that the library situation is bad - its not the lack of libraries, its the number of half-hearted libraries vs. the real good ones.
Also: Pro-active teaching, like producing screencasts and tutorials is sadly not very common in the Haskell community as well (weirdly, this is spotty: haskell-cafe, Real World Haskell and Learn you a Haskell are shining examples of how things should be, also in other communities).
I am surprised by how much the Haskell community always says "type system" first, but seems to put all other practical considerations on the backseat. Also, it is common to just dismiss concerns without addressing them.
That said, I still love Haskell and enjoy developing in it when things are working smoothly, I just wished that the community were more receptive and less dismissive to actual problems of people that are not as deep inside the language as people spending every evening on it.