Since the computer equipment is a "server that is set up in the facility" in Canada, it was used predominantly outside the US as defined under §1.48-1(g)(1)(i) for purposes of §168 from the get-go.  Why was it treated otherwise to start with?

If the server was located in the US and was subsequently relocated to Canada (assuming it does not trigger PE in Canada), it would constitute a change in use and require an adjustment to how the asset should be depreciated.  You may refer to §1.168(i)-4(d) for the details and Example 3, in particular, which clearly lays out the exact treatment.