It can also be work with too many grotty doesn't-teach-you-anything-technical challenges. For example, a bank that has acquired multiple companies with incompatible, archaic, poorly maintained systems, and struggles to keep them interoperating while stitching them together under some ill-considered technical policies and bizarre regulatory side constraints. Struggling with medium-scale technical challenges that only exist because larger-scale technical best practices aren't followed isn't as productive as it could be. You do learn things but it's akin to learning how to develop code in a dysfunctional language in the absence of specifications and source code control systems and testing: too many of the things you learn tend to be kinda messed up. And by the Anna Karenina principle they may not generalize well to your next job, even if that job also has more than its share of screwed-up challenges.
All businesses tend to have some proportion of work that is too idiosyncratic and ill-conceived to be satisfying or to be a good long-term career choice, and it's dangerous to overgeneralize from anecdotes. But I don't know any way to sharpen up a claim about the proportion of idiosyncratic ill-conceived work so that we could test the claim with proper statistical sampling. Falling back to anecdotal evidence then because it's all I have, there might really be a tendency for some parts of banking have more than their share of this.
Some banks can be so weirdly huge, that one department may be developing a simple website with last decade's technology, while another has computer scientists working on algorithms for a high frequency trading system with microsecond accuracy, which may never at any point have an unscheduled hiccup of a fraction of a second. As far as challenges are concerned, it's hit or miss, just like in any other sector.