The issues were related with the linux bridge on a specific compute node. This caused an error state which ended up throwing DHCP requests in a loop causing big spikes in traffic and latency. This already happened last night but the root cause was found this morning. Our cloud provider is tightening up our monitoring to detect such instances earlier in the future and we’re taking the necessary steps to prevent this from happening again. Our apologies for the inconvenience this may have caused.