Has the OP actually asked IB? If true, then cancelling the children before an acknowledgement that the parent has been successfully cancelled would be a bug in the design and would need to be fixed. I would be surprised if it was done like, however it would be trivial to fix as the children are not even live yet.
This is a separate point to what the OP was asking. Which IB cover with the 'over fill protection' flag. They will not 'eat the loss' on your behalf when you do not set this flag.
yes, see official ib reply: when you submit a cancel on any OCA (one cancels all) group (Child/parent orders would be considered an OCA group), the system attempts to cancels all the orders at once. We don't attempt to assume we know what you would want done if one of the order executes before the cancel reaches it. In some cases other orders in the group may be risk reducing but that will not always be the case so we attempt to cancels the entire group in lieu of making decisions for you on what orders to leave working. Given the parent order is in the market and child is most likely on our servers, it makes sense that in the view of milliseconds, a parent could be executed after a cancel submission. Cancels would have be sent to all orders in the group so the situation you describe is the result. Regards, IB
I'm using IB for "HFT" as well... Started with IB years ago, long before my systems were trading intra-day. I've heard many times that IB is "just barely ok" for what I'm doing... would be really interested in hearing of some alternatives...