Newer kernels aren't always better

I’ve been running the 4.19 LTS on a few production hosts for months now and it’s still rock‑solid. Every time I see a thread about "you need to update to 6.x for performance," I pause. Kernel developers throw in new drivers, patch bugs, but they also introduce regressions that hit the same workloads I care about.

If your hardware is stable and your software stack is fine with the drivers you already have, the extra mileage from the latest releases can be marginal at best. In my case, a 0.1% drop in CPU time during a nightly backup is a more tangible issue than a theoretical speed boost that only shows up in a microbenchmark.

I’m not saying ignore kernel updates entirely, but don’t treat every new major release as a must‑install. Stick to an LTS, test your specific use case, and only upgrade when you actually need a driver or a security fix. Let’s not get swept up in the hype—sometimes the old is the new.

What do you think? Has the latest kernel improved your workflow or caused headaches? I'd love to hear the practical outcomes, not the marketing spin.