Sounds like an issue of time preference. I personally prefer to not have checkpoints, doesn't matter if it takes weeks to go through multiple chains to find the one with the most work. At the same time, if a client with a checkpoint comes out, we got many node runners who don't change clients (and yes I'm calling it changing clients rather than "upgrade") for the first few years that client is out.

As we know, a new client that disagrees with old clients gets forked off and everybody is going to know about it.

So you know, a lot of nuance, but I personally run the more intensive options.