In other words, "success" (popularity, widespread use in industry, teaching, etc.) is not a justification for "costs" such as limiting expressiveness, making things less safe, restricting research possibilities, etc. Haskell would prefer to be powerful, safe, efficient, obscure and niche; rather than popular, industry-standard, widely-known, highly compatible, unsafe, insecure, inefficient and restricted (besides, we have loads of languages of this sort!).
Crippling the language to make things easier for new users is not an option. However, there are many ways the language could be made easier without impacting any core values. For example, there are long-standing issues with strings, records, error messages, dependency handling, etc. which people are working on.
Also, for those who do want to learn Haskell, I've found the community to be very welcoming and tolerant of newbies. Just ignore some of the more math-heavy blog posts (for now, at least), or your head might explode :P