There was a subtle true/false reversal mistake that we didn't catch when bringing in the new error enums. This was causing the second connection request to cancel the first ongoing connection request.