Consider e.g. a grumpy guy. Would it be good if he were to be diagnosed with the illness of grumpiness (with associated social costs) and prescribed a pill for that?

Well, among other things, it depends on what the pill does. If it doesn't make him less grumpy, for example, then I can't see any benefit at all. If it makes him less grumpy but also does a bunch of other stuff that leaves him worse off, then there's no net benefit. Etc.

All that aside, if being prescribed a pill that makes me less grumpy inevitably subjects me to social attack, I would say that being able to be less grumpy is a good thing, and the social attacks are bad things, and the net value of being prescribed the pill depends on the ratio of costs to benefits.

And in both cases, I would strongly endorse a social shift that stops attacking me for being prescribed such a pill, rather than blame the pill for the social attacks. There's all kinds of things I can do that subject me to social attacks; blaming me for doing them on that basis is suboptimal.

It assumes biological causality for what are not necessarily problems of human biology (or biochemistry).

Here again, it depends on what the pill does. If I start with a false theory of the causes of the problem, I'm unlikely to come up with a pill that actually solves the problem in an acceptably targetted way; by the same token, if the pill actually does make me less grumpy without too many side-effects, that's a pretty good sign that there's a biological (or biochemical) cause for grumpiness.

Incidentally, I would say all the same things about being prescribed a pill that makes me less cancerous.

The trend to consider certain conditions and psychological states "diseases" or "illnesses" (which implies biological causality) is bad because:

This is why I thought that tabooing "associated words" would be a good thing. Many treatable conditions in medicine are not considered "diseases" or "illnesses" anymore, and they shouldn't be. This especially applies to psychiatry. Many diagnoses cannot be made unless the condition "causes significant harm to the patient" is met.

I would go even further than most doctors, and say that it's a failure of medicine to only try to normalize harmful conditions instead of trying to improve upon what's considered normal. This means that the language of medicine has to change even further. A diagnosis certainly shouldn't automatically be a "disease" and not even a "disorder".

It narrows the range of what's considered acceptable human variation. Consider e.g. a grumpy guy. Would it be good if he were to be diagnosed with the illness of grumpiness (with associated social costs) and prescribed a pill for that?

"Acceptable human variation" and "illness of grumpiness" is a again a way to say there should be stigma attached to a diagnosis. I don't think there should and this is a separate problem from whether certain conditions should be considered treatable. If the guy thinks his grumpiness is a problem and causes significant harm, and it can be treated without side effects that are unacceptable to him, and he would knowingly accept the social costs, then I think he should have the option of treatment available. A diagnosis isn't an illness, it's a label that doctors use to communicate with each other.

Also keep in mind that therapy is a medical intervention too, and is usually better for specific behavioral problems.

It assumes biological causality for what are not necessarily problems of human biology (or biochemistry).

Like TheOtherDave said, the pill is unlikely to work, unless there is a biological mechanism involved.

There are considerable forces in the business world which would stand to gain huge amounts of money were this to happen. This is not an outright argument against per se, but it does make one suspicious.

In this case one should be suspicious of all treatments and not just psychiatric ones, and perhaps one should. I think this is a separate problem from whether certain conditions should be considered treatable. There's certainly a need for a system that has less perverse incentives.