System verilog randomization: inconsistency in LRM?

In reply to Tudor Timi:

That works in some cases, but you could not rely on it. Also, some people depend on this behavior to limit the scope of overriding soft constraints