Anyway, I don't think the global switch is useless. In the case of the other issue, the OP would have had to declare the tolerance on each of his constraints. So the global switch is more appropriate for him.

So basically, the global switch made sense and was easy to add.

Per constraint switches also make sense but, as Gunnar mentioned it, we either would have to create new auxiliary annotations (that would be doable, I drafted something in https://github.com/hibernate/hibernate-validator/pull/861#issuecomment-339984807) or we could wait for BV 2.1.

For now, we have decided to wait. But we might consider the other option if there is enough traction.