I have a "home" network with a number of Mikrotik components. On my property are 4 APs, all running 7.1.x. One of these is an Audience (running 7.1.3) that is running Capsman and controlling the others. The main router is a CRS125. I also supply internet to some of my distant neighbor (in a rural area) via some SXT PTP and PTMP wireless links. (These and the main router are running routerOS 6, not 7). All devices are connected with ethernet to the CRS (except for the remote neighbors, obviously). All was working properly until I added an hAP ac3 to the network for testing. I configure the hAP ac3 (running 7.1.3) as a switch/AP using the Quickset option for CPE after resetting the device config to defaults. I assign it a unique static address on my network and turn off dhcp on the hAP.

As soon as I connect the hAP to the network, the network becomes unstable, seemingly dropping layer 2 packets, A ping plot to my remote neighbors show that no traffic gets through. If I disconnect the hAP ac3 the network stabilizes in about a minute. No errors show in any logs. After trial and error, I discovered that if I disconnect my Audience (running capsman for the APs on the local campus only) the network works find with the hAP ac3 connected. I conclude that this has something to do with the Audience. If I disable Capsman manager and ipv6 on the Audience, the network still becomes unstable when both the Audience and hAP are connected. If either is connected by themselves, no problems are visible.

Any suggestions on what is causing this or how to debug it would be appreciated.

Further testing allowed me to isolate, but not yet solve the problem. If I connect that hAP ac3 directly to port 2 of the Audience (instead of connecting it to the CRS125 where all other equipment is connected), the problem becomes isolated to only the link between the Audience and hAP and does not impact the larger network. I've check mac address and there are no overlaps (or errors about duplicate macs). Also pings between the Audience and the CRS125 become unreliable for minutes at a time and then go back to normal. When failing, every other ping from the audience to the CRS fails, though the CPU times on both the audience and CRS are less than 5%. The link between the CRS and hAP is a 3 foot ethernet cable. The status shows no errors from either end, running at 1 Gbit.

First thought on L2 instability .... (R)STP : spanning tree transitions, root bridge elections.
Is the root bridge where you want it? Set low priority number on that one (e.g. 7000). L2-network diameter? (=longest path of bridges/switches).
SXT PtP and PtMP a lot of bridges there. Any modification at an endpoint with a (R)STP capable switch or bridge will make the network calculate the spanning tree. (even enduser devices). During that time transmissions stop.

@bpwl: You were right on. When I looked at STP settings on the main router, for some reason I discovered it was set to "none". Changed it to RSTP with a Priority of 7000 and all works. Not sure when or how STP got turned off.