> but in practice the opposite often turns out to be true
One backwards compatible issue in how many years, builds and packages? I would say _often_ is highly exaggerated. I've had many more issues with rolling release distros than I've ever had with CentOS, I can basically do a yum upgrade and reboot with my eyes closed and not have any issues.
One issue that eats your data is worse than 100 issues that just take up your time to fix. And sure, you can say backups are important, but we know that in practice, a lot of people don't keep good backups.
And besides, it isn't just one issue. https://bugzilla.redhat.com/show_bug.cgi?id=2039993 is another example of a problem caused by incorrectly cherry-picking a security patch: OpenSSL was affected by CVE-2021-3712, which they quickly and correctly fixed upstream. But RHEL still uses OpenSSL 1.0.2, which has been EOL for more than 2 years, so Red Hat had to backport the patch themselves. When they did so, they made a silly typo (forgot the "!"), which made Web servers all crash with a double free error very shortly after every startup.
https://bugzilla.redhat.com/show_bug.cgi?id=1861977 was yet another, in which trying to fix a Secure Boot-related vulnerability rendered computers unbootable. A fourth one, which I unfortunately don't remember the bug ID for, was when a patch added to NSS in a security update made programs hang when you tried to use a smart card.
To be clear, I didn't go searching for these problems. I know about them because I was affected by them all.
OpenSSL 1.0.2 is a library used by third party software. the value proposition of SLES/RHEL/ubuntu LTS is binary compatible drop in replacements so you don't have to rebuild your user land.
Firefox is a run time environment for the UI for POS/kiosk systems. With horrific customer applications running on top. I can only imagine the internal discussions inside these distros wether to upgrade firefox/chromium or wether to backport patches.
That's why each of the profesionally backed distributions has some strategy to mix the frozen version maintained baseline with some more recent subsystems "from the future", SUSE calls it "Leap", for example.
So I'm not sure if you suggest to completely abandon frozen version supported platforms? or do you suggest improving upon the model? so fewer issues like the ones you listed occur? "even fewer" I should say, because the events are rare. painful, but rare.
One backwards compatible issue in how many years, builds and packages? I would say _often_ is highly exaggerated. I've had many more issues with rolling release distros than I've ever had with CentOS, I can basically do a yum upgrade and reboot with my eyes closed and not have any issues.