Can you reproduce this?

This happened more than once (same question in the same questionaire). I would say it's reproduceable.

Can you give us any hints on how you did it?

You're not having multiple editors working on the same survey simultaneously, right?

Hi,
no, not using multiple editors. Just one browser window (latest Vivaldi).
I was thinking that caching could be the part of the problem? I am developing databases and it's so strange to me, that a value, once saved to the database is changed when I edit something completely different.

The problem here is the bad distinction between customized ExpressionScript and using the conditions designer.
In LS3 you could either use condition designer OR edit the relevance field manually.

Now you can use the condition designer AND edit the field afterwards - the condition however is not removed and at some point the conditions get translated to ExpressionScript again, overwriting the customization.

How it should work:

If a conditions exists and the user starts editing the condition directly (not using the condition editor), a warning message will be shown:
"Note: If you customize & save this condition you will not be able to use the condition editor for this, anymore."

If the conditions is customized and the question is saved, the condition data needs to be automatically removed.

A similar note should be shown if you use the condition editor and a customized condition already exists.

"Note: This question uses a customized condition. If you create a condition using this editor the current customized condition will be overwritten. "

@ong: To prevent that the field get overwritten again, backup your custom condition (copy/paste to somewhere) remove the conditions from the editor first, then save your custom condition, again.

@c_schmitz
Yes, that sounds good, but my original problem why I did edit the condition by hand was, that I was unable to find a way to connect the conditions with the OR operator. Using the condition designer the operator was always AND.
Maybe there should be an easier way to use boolean operators to connect conditions in the condition designer.

The underlying issue will be resolved in 16462 so it behaves like in LS3.