To be fair to the idea, though, while this would make individual instances less secure, it would drastically decrease the leverage for the work bad actors put in.
There is a saying in the software security industry that (I'm paraphrasing from rusty memories) a system is secure if the cost of hacking it is higher than the value it protects.
Each system being completely distinct from another means that the cost of hacking the average student goes up by 9000 (from the article, Canvas is used by 9000 schools).
Still not saying that rolling out your own is the preferred solution, but the idea is not as ludicrous as it would seem, and should definitely be entertained and discussed, at least.
Put it another way; the blast radius from any vulnerability is much smaller.
But also, the cost is much, much higher to the institutions, which is the salient point. You're going to spend years developing a system, deploying it, training staff and students, supporting it. I see mentions here of in-house systems being developed much more cheaply and I don't believe it. The economies of scale are at work.
I worked at a university for many years and I can't recall anyone I'd consider to be a competent software architect working for the IT department. Hell, we had students writing major webapps that kinda sorta worked well enough.