Think about your second sentence a bit, and why it's true: major Linux distributions adopted systemd. Did they do that because it really caused a bunch of trouble or because a deliberative review process showed that it solved a number of hard problems?
A small percentage of people having very loud emotional reactions doesn't mean that it actually caused large numbers of problems and at least in the enterprise space it's been really useful for cleaning up long-running reliability and security problems by removing thousands of lines of SysV kludge and manual work recovering from problems.
But my point is a different one: If you manage such a project, you have to manage the change. And when you know that there are people who disagree with you on a fundamental level, it doesn't help driving your followers away by releasing breaking changes on a regular basis.
As an end-user, I remember for PulseAudio and systemd at least one instance where I had built something with it and after an update, it didn't work anymore and I had to adapt it.
So I don't believe that there was just a 'small percentage of people having very loud emotional reactions'. Instead, I think it was more like everybody had some problems adapting to systemd, but some saw the benefits it came with and others fought it as hard as they could (for various reasons).
So the problem isn't what Lennart had done (which is great), but how he got there. And I think Lennart would be happier today if he didn't do it the way he did.
It is hard for me to imagine what it could be doing with so much RAM.