Maintenance to a disabled compute node caused an unexpected packet storm on a network switch impacting legit traffic between instances on other compute nodes. We have implemented a temporary fix until we can patch the network switches which caused the issue.