How much variance is acceptable in a bed mesh?

A range of about 0.02mm corner to corner is genuinely good. Most people would be thrilled with that. As a rule of thumb, anything under 0.1mm total variation is considered fine for everyday printing, and you are five times better than that. The gentle bowl shape is also normal, since beds tend to sag slightly in the middle from the heat and the mounting points at the edges. What you want to avoid is sharp spikes or dips between neighbors, and you do not have any of that. Here is the part that matters most: your first layers are sticking and you have no failed prints. That is the actual test. A mesh is a compensation map, not a report card. Klipper is using those numbers to move the nozzle up and down as it prints, so a small bowl gets handled automatically. Chasing a perfectly flat mesh on a textured PEI plate is mostly a hobby in itself, expecially since the plate itself has texture that hides tiny variations anyway. One thing I would check instead of adjusting screws: run the bed mesh command twice in a row and compare the two maps. If the numbers shift around noticeably between probes, that is probe noise, and no amount of tweaking will fix it. If the maps look nearly identical, your probe is repeatable and the mesh you have is trustworthy. Also keep in mind the mesh was taken at one temperature. A bed at 60 degrees for PLA moves a little compared to cold, so some of what you see is just the plate doing what plates do when heated. If it were me, I would save the mesh, stop looking at it, and go print something. Revisit it only if you start seeing first layer problems in specific spots, like squished lines on one side or lifted corners. Numbers that pretty with successful prints are a sign to walk away, not a sign to keep turning knobs.

Totally agree with walking away here. My own mesh used to stress me out til I put a dial indicator on the bed one weekend and realized half the variation I was seeing was the probe itself, not the plate. Once I knew my readings had a bit of slop in them I stopped obsessing. Also worth checking the magnets or mounting screws are snug if you do want to tweak, a slightly loose plate corner caused a low spot on mine that showed up in every mesh until I snugged it down. But with sticking first layers I'd just print and forget it.

One thing I learned the hard way with my mesh: make sure you are actually saving and loading it in the right place, because Klipper will happily use a stale or default mesh if your bed_mesh section is not set up right. Mine was printing fine for weeks and then I checked the startup log and it had loaded nothing at all. Also worth knowing that the mesh only compensates within the area you probed, so if you print something near the edges past your probe points it just extrapolates, which is where you get wierd squish lines. Check your min/max bounds match what you actually print on.

Yeah that matches what I see on my machine too. I ran the mesh twice like suggested and my two maps were basially twins, so I stopped touching it. One extra thing that helped me: I take a fresh mesh whenever I swap nozzle sizes or change the plate, since a 0.6 nozzle or a diffrent plate sits a hair differently and the old numbers stop matching. Took me a while to figure out why my first layer got sketchy after a nozzle swap when nothing else changed.

So, I checked the setup and made sure the mesh was saving and loading correctly in the software. That helped stabilize the numbers a bit. Still, I'm looking at that gentle bowl shape and wondering if I should keep tweaking it, or if 0.02mm variation is genuinely good enough.