Any supporting arguments for 1-10 = 1st decade other than technical perfection? I guess if you use data around and before 1AD you care about this more, and rightly so, but given sound arguments for both methods the one more useful to more users who I suspect dominantly care about years > 1900.
So -1 to change for breaking backward compatibility and -1 because the current behavior seems to be more useful in everyday usage.
Since there was no year zero: then it follows that the first decade comprises years 1 to 10, and the current Millennium started in 2001 - or am I being too logical??? :-)
This is SQL, only relational logic matters. All other logic can be superseded by committee consensus.
IOW - and while I have no way of checking - this seems like something that may be governed by the SQL standard...in which case adherence to that would trump mathematical logic.
