Strictly speaking, isn't the CNC program the design
of the process used to create the robot part, rather than necessarily an element of the design of the part itself? I grant that the design of the part and the design of the CNC program might certainly be codependent, but fundamentally, they're separate things. Completion of one does not necessarily imply completion of the other—it would be quite reasonable to contend that the design of the part depends on the lessons learned from the implementation of the CNC program, and that the "final design" of the part is incomplete until the results of a trial run are analyzed and deemed satisfactory.
And although FIRST hasn't said so explicitly, I give serious credence to the idea that the robot rules must be assumed to only apply to actual parts of the robot, unless otherwise specified. (If you read the LabView example as applying to non-robot code, it opens up a world of unenforceable insanity.)
Put another way, logically, if you were to instantiate the part design, you'd get the part. But if you instantiate the CNC program, you get a machine following a toolpath—and maybe a part as well, but only if you decided to cut metal (or whatever the part is made of) rather than thin air.
And in any event, like Chris and Cory said: if you're still uncomfortable, make any one tiny geometric change, and it's unambiguously no longer the same part. (The old part can rightly be called a prototype, if you happened to make any. Don't use the prototypes in competition.)
Following my own chain of reasoning, I disagree that posting it publicly would be a remedy under the rules—stuff that isn't part of the robot has no particular reason to be (or to become) COTS.
Aside: See here....