I would think your first couple points are actually better on a merged system - if /bin==/usr/bin and /lib==/lib64==/usr/lib then scripts can use either /bin/bash or /usr/bin/bash and it'll work (and ditto for libraries). Granted, removing those symlinks would then be painful, but I haven't heard of that being proposed; if it has been and I'm just out of the loop then yeah that'd make me nervous. (Though in fairness, scripts really should use `/usr/bin/env bash` for portability anyways - /bin/bash was never more than a distro-specific quirk.)

As to testing... yeah obviously this stuff should be thoroughly tested long before it hits users, but I was rather under the impression that it was tested extensively before getting to users. Again, if you've seen actual problems that made it past testing, feel free to point them out since that would greatly strengthen your criticism.

> I just want an OS that works without having to fight someone every 3 weeks because "everyone else is doing it, so we should too".

Are you super committed to Linux OSs? Because you might find the one of the BSDs or illumos more comfortable. Not that they never change (er, well, illumos might not), but at least it's not because of what anyone else does.

(And as an aside: Don't take this as an endorsement of merged usr; I actually prefer a separate /usr myself because I lean towards thinking that it's reasonable to have a read-only root and a rw /usr, or /usr on shared NFS, or any number of other "weird" systems... but I also culturally favor the BSDs, so take that as you will. I just think that if we're going to find fault, we should have the best evidence/arguments possible.)