From the article, the subpoenas went to Google, Meta, Reddit, and Discord, and are targeting accounts that "lack real names".
Even if someone has signed up with fake names, I hold tremendous suspicion they can ID people anyway. Of course their accuracy is going to depend on many things. For example, did someone use their own IP, or a VPN? Did they use the app, or a web interface? But some of those sites won't even allow signups through VPNs.
There is a big industry built around browser and device fingerprinting for "identity resolution". This tech was driven by the advertising industry, after their ability to use cookies for identification was curtailed. I have no way to prove this, nor evidence for this, but I would eat my hiking boots if the big social media sites did NOT use that sorta tech.