I've previously explained my long-standing scepticism of clean definition classes along smooth, continuously varying traits that aren't multimodal. Imagining something is on a bell curve (many things are), where do you draw the line?
This complaint was recently echoed in Scott Alexander (actual psychiatrist) saying the same thing about ADHD – that it didn't exist as a cluster, but simply was the tail of a distribution of attention (I'm paraphrasing, obviously).
It's worth saying that there is a relatively simple way to get a fair guess at the answer to the question, though: find all diseases where diagnoses seem to steadily climb across at least one order of magnitude, and ask if they can be explained by population traits shifting over time. I think it's fairly clear Generalized Anxiety Disorder is an example. Something is clearly wrong! But not discontinuously wrong, or even necessarily multimodal.
I think it's probably good to normalize people trying to be self-aware, and using that as the beginning of the process of trying to be better. While I'm a bit concerned about hyper-medicalizing these things (the population trends are happening for non-medical reasons!), I think medicines are a tool, and many times the best options. Obviously, it's good to improve things and bad to not improve them, so any societal reticence to engage here isn't great.
The obvious second example is intellectual disability, which is literally defined as below 70 IQ, essentially being a perfect example of the strange clustering and pathologizing I think isn't helpful on mono-modal, standard distributions. It isn't good to be stupid – and we should help people be smarter – but nobody's perfect here, and 'smart' isn't a moral characteristic. It's just an axis we'd prefer to move along. Let's take the time to imagine these things a little bit more charitably.