Having worked for both Sq. D and Siemens, I can attest to the fact that the breaker CANNOT move from On to Off on its own, doing so would violate NEMA design standards. Even with a shunt trip or under voltage release, the device operates the TRIP bar, so the breaker moves to the Tripped (center) position, not Off. Even the mechanical force of opening a dead short at the maximum interrupt capacity of the breaker is not allowed to over travel the handle mechanism to the full Off position. There must be a distinctly different Tripped position.
It's possible that there was a defective breaker. Siemens had a recall a decade ago on some 400A frame Sentron breakers because of a similar issue, having to do with a small plastic part in the handle that was out of spec and could break off, which could interfere with the proper handle operation, one possible outcome of which was reported to be the lack of a trip position. But 3 defective breakers with the exact same defect that Sq. D is unaware of? Not likely.
POSSIBLY there is a motor operator on this breaker that the OP is unaware of. It might be tied to some sort of energy or power quality monitoring system and when it sees something going wrong, is MOTORING the main breaker open on purpose. Something like a Sq. D Powerlink MVP system. For all we know, this could be completely normal, the OP has never reported back a part number for the breaker.