There's this really frustrating dynamic where "lots of single-purpose libraries" is very clearly the optimal model once you do a full technical analysis, but it doesn't *feel* that way.

Because what people remember of this model is "having to wade through hundreds of packages to find what I need and getting list".

Which wouldn't be a problem in a correct implementation of this model (where you provide a pre-bundled/pre-configured collection for the common case and clear documentation on what you need) - but in a correct implementation, most people would never interact with the underlying hundred libraries, and so they will never perceive it as an example of the "many small libraries" model to begin with!

Do it wrong and it'll be a bad example of the model, do it right and it won't be seen as an example of the model at all. Damned if you do, damned if you don't.

(See: npm, OpenStack, ...)