I just power cycled both APs.  One had no clients connected and I could not force one on to it.  After reboot, clients balanced between them and are connected to the best AP based on signal now.

I did take the troubleshooting logs from both before rebooting so if someone from TP-Link wants to reach out to me I can provide those logs.

  @Vincent-TP 

I have a total of 36 clients.  

5 total SSIDs.  Here they are (names changed to protect the innocent 

Main   (2.4/5ghz both Dynamic PSK)

Main-2g (2.4ghz only Dynamic PSK)

Main-5g (5ghz only Dynamic PSK)

Special (2.4/5ghz assigned VLAN)

Guest (2.4/ghz assigned VLAN, guest network with client isolation)

  @Vincent-TP 

Range of clients.  PCs, Mobile, Android Streaming, IoT (bulbs, switches, plugs).  Either way, 36 clients is not much for these.  And I never had problems with this up until maybe 4 months ago.

If I do WAN optimization it does not do well and puts both APs on the same 5ghz channel (I have seen posts where it does this to others as well) when there is no need for it.  Not enough interference from neighbors to require it.  So I have them manually set to be on difference channels with the lowest usage.

FYI, Woke up this morning and one failed last night when there was very little overal usage and activity.    Here are the first couple graphs from both.

At 3am, the client count on the masterbedroomap dropped from 16 to 2 and the officeap went from 18 to 34.

At 3:10am things swapped some, officeap dropped from 34 to 5 and masterbedroomap went from 2 up to 30.

At 4:35am there was a spike in CPU usage on masterbedroomap and client counts started to drop eventually reaching 0 around 6:05am.  CPU never dropped even when this AP hit 0 clients.

By 6:05am officeap was now handling all clients.

And while not on this graph, as I write this, at 8:30am CPU on officeap is now spiking and clients are dropping from it.  Time to reset them both (which I do with a full power cycle). 

Here is an interesting observation.

When masterbedroomap started experiencing issues at 3:00am.  There was a spike of mesh traffic at that AP.  I have no other APs and both of my APs are wired.  Mesh is in fact disabled on my controller even.

I just factory reset both APs and re-adopted them.  I then ran AI WLAN Optimization and I'll leave what it recommended.

One is on chanel 36 and the other on 40 for 5ghz with 80mhz channel width.  Both were set to 24db transmit power.

However, it set both to Channel 6 for 2.4ghz.  This just seems dumb to me.  While 6 is overall the lowest use in this area, having both on 6 and also at max power makes no sense.  Channel 11 is used less than channel 1 by neighbors so I would have expected it to use channel 6 and 11.

  @Vincent-TP 

So far so good but it's only been about 12 hours.  Even with both APs on channel 6 for the 2.4ghz band I don't "seem" to be having issues but most of those devices are low bandwidth IoT devices.  I'll let things go with these settings for a week and compare 2.4ghz packet loss and errors to the previous settings when I had channel 1 and 11.  The bigger goal is to resolve the APs tendancy to drop all client connections and require the reboots.