System verilog randomization: inconsistency in LRM?

In reply to dave_59:

Some tools work back in this case and choose a different value for y, such that the randomize doesn’t fail.