Yes, I think this is new behavior. I have recently changed the code around what happens to unbeveled edges attached to a vertex where at least one other edge is beveled. It seems different choices are being made at each end of that beveled edge.

Interestingly, from reading Tomas's initial description and Bastien's last comment, it may not be immediately obvious which end is "right", but it certainly looks strange when there are different choices on the different ends. Seems like Tomas prefers the left side (no corner bevel) and Bastien prefers the right (corner bevel). From recent artist feedback, I was going more for the effect of the left side (no corner bevel here).

I will dig into what is going on here, and try to make both ends look like the left end (no corner bevel).

Hi Howard, as you wrote it is not obvious what result is right. I would like to say that the "left one" is only my personal preference. This is something what I would expect from my experience.

Anyway I was checking other 3d packages and results are also different but mostly "right one".

We discussed this with my coleagues and the option to pick result would be the best. But you know us artists we want evrything ;)

Nevertheles I agree with you that result can be any of them but same for both sides.

Thanks for the bug report.
It turns out that both ends were creating a polygon around the ends of the beveled edge, but I was using an inconsistent way of determining the length to 'slide' along unbeveled edges. The commit in the previous comment fixes that.

Another possibility would be to move further in the direction of letting 'in-plane' edges move all over the place and use that to avoid creating vertex polygons at all here. The result would look like:

But that requires more special casing in the code, so will not do that unless more demand for this crops up.