Overall I think it was well thought-out. I have a few comments:
1. My primary question is: How are they going to enforce this?
It's not illegal for women to have high testosterone levels. It's only illegal for them to have high testosterone levels if they have DSD. So what is the protocol going to be for the next Caster Semenya? If drug tests show a new runner has >5.0 nmol/L and she has no external male genetalia, then she's obliged to undergo some sort of doctor's exam to look for internal testes? If the doctor's exam is inconclusive then they have to do genetic testing to make sure there's no chromosomal abnormality? IIRC, part of the problem with the Caster Semenya case was that they didn't really inform her of what they were doing.
2. As a secondary comment, while it's clear now that Caster isn't the only athlete who will be affected, it does seem kind of obvious that she was somehow targeted, since they targeted only the 400, 800, and 1500/mile (plus 400 hurdles) -- they made sure to shut her down for her whole range of competitive events and nothing else. If they care so much about this for some sort of abstract reasons of fairness, why *didn't* they go after shorter events?
3. Similarly, something bothers me about the justification of this as a performance-based decision.
First, Amby makes many comments about how her "peers who strive as hard as she does but have no chance of winning". But it's not true. She hasn't destroyed the all-time records. She hasn't beaten Pamela Jelimo's time, even though she ran a decade later. Her PR is less than half a second faster than Ajee Wilson's. Ignoring Niyonsaba, the next fastest runner of the decade after Ajee Wilson is Natoya Goule, who ran half a second slower than Ajee Wilson, and the next fastest runner is Habitam Alemu, who ran over a second slower than Ajee Wilson (according to alltime-athletics.com). So Caster Semenya is just as much of an outlier as Ajee Wilson would be without Semenya and Niyonsaba. Just to put things in perspective here. [Related side comment: What's really remarkable about Semenya isn't her PR, but her consistency. Semenya has 3 of the top 10 performances of all time, which is more than anyone else. Pamela Jelimo has 5 of the top 20, but Caster has 8 of the top 30.]
Second, this is only a concern in the first place because Caster Semenya started winning. You can't know that the fact that she's winning and has a DSD implies that the DSD is what causes her to win. There's a phenomenon/fallacy (in statistical genetics) called "Winner's Curse" that we should watch out for -- the first discovery of something tends to outshine subsequent examples of it. Regarding the increased rate of DSD athletes winning medals, there's also a phenomenon that small sample sizes have large error bars. "DSD medals in the women’s “restricted events” at the World Championships and Olympics during the last 25 years" -- I don't have access to the journal article, but doesn't Semenya herself account for like half of these? And there are like two or three of these athletes in total?
Third, maybe she IS doping. Maybe all of these DSD athletes are doping.
Fourth, this quote: "In Lausanne, the IAAF team presumably submitted additional data linking testosterone and real track times." I don't think anyone doubts that the results of such a study would be huge -- most of us on here know there's a roughly 10% difference between men and women in most of T&F, and many of us wouldn't be too surprised to hear that testosterone works. But it's still not clear to me how someone with androgen insensitivity syndrome is going to be affected by an increase in testosterone levels. Isn't the whole cause of Caster Semenya's unusual development due to the fact that she was unable to make use of the excessive amounts of testosterone in her system? It doesn't sound to me like this establishes that she has an advantage. So what's the point? She might have an advantage, and we're not sure, so we might as well try to limit her testosterone levels?
I'm not saying this isn't a good decision, I'm just saying that there's something about this sort of justification that rubs me the wrong way as a statistician.