IP level temporarily disconnecting clients

I have one rb433 as AP with two wireless MT-clients.
AP wlan interface is with ehter interface in bridge. Bridge has the IP set and is gateway and runs dhcp server.

Changed sub netting last day but since then both clients loose their IP level connectivity repeatedly but irregular every so many minutes, to come back after some 10 -40 secs… It happens to them at the same time. (This makes me think something with the AP?)

Clients are rb113c units.
All three units run ros 3.28 (and tried 3.27, same issue)
Clients are dhcp client from AP which is server
All three units run for two years without similar problems. Up to last night everything fine in a /24 network. But since last night changed to /27 network.
/27 on the AP as gateway, as dhcp server (ending .65/27) and routers higher up pointing to this /27 network. Old routes removed and checked on duplicate ip addressen anywhere else on the network. (so no IP conflict althought the problems looks like one)

Both logs in AP and CPE show nothing.
Running mac ping from AP to both CPE’s never broken.
Running IP ping from AP to both CPE’s broken each time, to always come back after 20-60 secs.

I checked routes and sub netting. When units are up and reachable by winbox. I can ping and traceroute anywhere to my network and the internet. From my central I can reach them by ping, traceroute and winbox when they are up.
This basically means that routing is all set up fine.

Further more, changed similar network somewhere else exactly the same but with 8 MT units. No problems of the same kind.

I am puzzled here, anybody with some suggestions…?

Ok, when I disable ether1 interface (which has no active devices attached) the link to the two wireless CPE’s becomes stable for 15 mins at least. Enable ether1 again and the issue is back after 4 mins…

So, took the ether1 out the bridge now and things seems to be fine.

Still puzzled why this problem. Other AP’s run fine with ether1 (which is POE and always has at least one client on it too) in the bridge.
IP firewall option for bridge not enabled sine their is always one other wlan as backhaul that is routed. Thus normal firewall takes care of everything.

So, it looks like an IP conflict coming from the ether1 network. But apart from a small indoor wifi router no other units attached?

Any other suggestions…?