If EC_CMD_REBOOT_EC goes wrong (eg. we don't actually sysjump because we're already in the image we're asked to sysjump to) then we'll send an extra protocol response -- one early response from the call to host_send_response(), and one response after we return. Eg.

# ectool reboot_ec RW
ioctl -1, errno 90 (Message too long), EC result 0 (SUCCESS)

Maybe we should consider adding a return value or state flag that tells the host command task that we've already sent a reply so it won't send another one.