Hi Mr. Yuri, I've been able to track down the bug which was causing this - it is in boolean union where it is attempting to merge 2 cylinder surfaces into single larger cylinder but is making a wrong result when the 2 cylinders have opposite axis directions.

I've been trying to find this bug for quite a long time without success until now, I had thought it was a bug in filleting since all the cases where I've seen it have been on fillets. But until now I've only had examples of it after it has already gotten messed up which made it difficult to figure out.

Thank you for posting the file especially with the "before" and "after" states, it helps a lot to be able to repeat the problem instead of only seeing the end result. It will be very good to get it fixed up.

- Michael