If someone did do that, I think we'd probably know about it. The chance of getting caught hacking a school district like that is tiny - but accumulate hundreds of tiny chances and detection is almost certain.
If you're good and/or committed enough to maintain hundreds of covert hacking operations, though, there are likely better options. How about deploying facial recognition at scale on the Internet? I hear they're pretty awful compared to humans at identifying people from limited data, such as a few grainy surveillance photos... but if you already know who you're looking for and have many high resolution photos of them, it would only take one good shot of them in their new life to be uploaded to the Internet - maybe in the background of someone else's photo, or even a frame of a nice HD video - and I suspect computers would be up to the task of making the match.
I am skeptical about the first part (the second part, about facial recognition, seems good, and at this point might just be a budget issue -- that to comb through the low quality results manually given current software would take too many man-hours).
I have seen IT for school districts. It's often one person, who has an insane number of demands on their time, no resources, and no specialized knowledge in security, building a number of systems that handle sensitive and non-sensitive data the district cares about. Given how homogeneous that situation is, evading detection consistently seems... extremely doable. And being caught 1% of the time means they have access to 99% of the networks -- they aren't magicians, there wouldn't be a clear way to intuit the larger context. If they all use e.g. web based grading software, and that gets 0-day'd, your 'almost certain' estimate gets readjusted by quite a lot too, and that's not a crazy thing to have happen either.
If you're good and/or committed enough to maintain hundreds of covert hacking operations, though, there are likely better options. How about deploying facial recognition at scale on the Internet? I hear they're pretty awful compared to humans at identifying people from limited data, such as a few grainy surveillance photos... but if you already know who you're looking for and have many high resolution photos of them, it would only take one good shot of them in their new life to be uploaded to the Internet - maybe in the background of someone else's photo, or even a frame of a nice HD video - and I suspect computers would be up to the task of making the match.