Refuse makes an excellent point on the other thread about the fact that disclosure of data from MN potentially risks physical harm to individuals.
The assumption in commercial DPA is that ultimately the concerns are financial. If it goes wrong, the risks are (a) that customers/etc will sustain financial losses and have to be compensated financial and (b) the ICO will fine you, which is a bottom line loss. It's money. In medical DP, the risks are more diffuse but people are (in general) assumed not to be harmed by disclosure in a way that cannot be dealt with by compensation, apology and so on.
If MN are indeed holding data which identifies people at risk of serious physical harm possibly including death, which seems very likely (at the very least, MN hold a data set which is capable of being analysed to produce data which puts people at risk of physical harm) then, in the old world of HMG Infosec Standard No. 1 ("IS1") MN are holding data whose disclosure risks "Permanently incapacitating injury or illness to a group of individuals impacting work and leisure activities. The individuals will suffer with permanent disabilities. Loss of life to an individual. Health and Safety Executive review undertaken due to the scale and impact of the crisis".
That's IL5, SECRET. That's heavy shit. Massively complex handling and management rules, change-controlled and audited infrastructure, everyone with routine access cleared to SC. It's hard to see how you could run a public forum where the aggregated data and the underlying authentication data is IL5. And arguing data is "UNCLASSIFIED, SECRET in bulk" is going to be a big ask.
Indeed, Refuse might could argue it's data capable of "Permanently incapacitating injury or illness to a large group of individuals impacting work and leisure activities. Potential for a limited loss of life. Health and Safety Executive enforcement notice leading to prosecution due to the scale and impact of the crisis", which is IL6, TOP SECRET: everyone who goes near it DV cleared, etc, etc. But it doesn't meet any of the other tests for TOP SECRET, not remotely, so let's settle on IL5.
It would ambitious, to say the least, to attempt to assure a system in which millions of uncleared users have access to a large web platform where parts of that data are IL5. I'd go so far as to say it's impossible, in any meaningful sense.
This is, in 2018, new and uncharted territory. I suggest MN might like to talk to some academic and commercial IA people in an NDA's environment to discuss options. But having interns able to see identifying data and exfiltrate it in an uncontrolled way isn't a promising start...
(Added on this thread: yes, you need to also consider the capability and motivations of the threat actors. Based on yesterday's incident, and her friends, I think it can be assumed that both capability and motivation are proved).