These are good points. I'll respond with a few things that I think are relevant:
> an argument for being a late-adopter
Definitely. I was an early adopter of 23andme, and there is value to 23andme's research arm when a customer fills out the surveys, but the utility is low and comes predominantly in the form of entertainment. I'm the most neanderthal of everyone in my family who has signed up, which is a mark of pride.
> A small increase over the average population susceptibility for trait X is often non-actionable for you personally, but over several traits might be sufficient to shift you into a different insurance risk class.
While this is quite true, and would be very worrisome, I think the basic properties of heredity make it unlikely that the dystopian outcome you describe will occur. Here's what I think is the logical argument:
We are all very genetically similar. There are thousands of 5th, 6th, 7th, and more distant cousins of mine on 23andme. Chances are most of the population descended from Europeans are Broad, ethnicity-based risks are already known/used by insurers... things like sickle-cell risk, diseases more common among Ashkenazi jews, etc.
While there are many health conditions that correlate with genetics, their presence is dominated by chance. Thus we all have similar incentive for adverse selection of insurance, and hence insurance companies have little to gain by fragmenting their risk pool on the basis of tests with low predictive value.
Moreover, since so many people reproduce and thus take on financial responsibility for the genes of their children (which are subject to both hereditary risk and mutation risk) chance becomes an even more dominant factor in risk, and amplifies the incentive to create the largest risk pool possible because the risks cannot be predicted with high levels of accuracy. There have already been extensive, large data set statistical analysis of full genomes for predictive information about the top 5 most costly diseases. If there was going to be a large, statistically significant find, we'd have already heard about it.
The other side of this coin is that even people whose genetic scan shows reduced risk for all of the expensive diseases will still get those diseases with enough frequency that segmenting the risk pool will not make financial sense for insurers.
The situation we had a few years ago with respect to expensive pre-existing conditions was actually far worse. Type 1 diabetes is weakly hereditary and is dominated by chance. If you get type 1 diabetes you're looking at a few hundred dollars per month of cost for test strips, insulin, etc., to the point where for a middle class or lower middle class person, the disease can be financially crippling. Most people would prefer that the risk of randomly occurring, expensive diseases be spread among everyone, so that we'd all pay $1 more so that the person who gets unlucky and has type 1 diabetes gets the strips for free.
Yet in spite of this, even with acknowledgement of pre-existing conditions, the scenario remains. The insurance/healthcare industry has not managed to create the proper financial incentives that would insulate someone who drew the short straw and got type 1 diabetes from financial hardship (via pure insurance)... while it has failed to create any financial disincentive for adopting many behaviors that are nearly guaranteed to increase healthcare costs.
So I have hope that a better (broader, more public) understanding of genetic risk, random chance, and behavioral risk will help the insurance industry deliver products that allow all of us to save money based on things we can control, and not to be penalized for things we cannot control.
Neither of the extreme ideas (that there is strong genetic determinism for disease and healthcare cost or that there is zero behavioral/environmental impact) are true.
Also, fwiw, I don't think any of us would tolerate living in a society where people had to consider the genetic scan of a potential mate before deciding whether a relationship made sense financially. I realize this is the case today in some minority populations who have had (for historical reasons) lots of in-breeding) but over time the impact of that will be reduced.