I have two EAP670 APs. For the last few months (since 1.0.10, if I recall) I have seen what looked like a spanning tree loop that would occur around weekly, but sometimes more frequently. It turns out this is not a spanning tree / bridge loop, but a multicast flooding issue. I have seen a number of threads pointing this out and providing possible solutions. I have gone through these solutions and tried, I believe, every one of them. I was finally able to capture a packet trace using tcpdump during one of the events. It does appear to be caused by a google/nest dot, but even removing the google device from the network during the event does not clear the issue. The only solution is to power off one of the APs and power it back on after a few seconds. This multicast loop is incredibly frustrating, and has made me consider move back to my EAP615-Wall APs. I don't know really what else to try, so if anyone has any ideas, I'd love to hear them. FWIW, the latest firmware (1.0.14) made the problem an order of magnitude worse. 

Interestingly, the packet trace shows that the google device which is on the SSID attached to VLAN4, which is tagged to the AP, is showing up on the untagged management vlan (809). Attached is a network diagram:

FWIW, I have also removed VRRP to reduce complexity and yielded no improvement, as expected.