I'm new to RBE970/971 but old to all this netz stuff

Well said; you’re right as a general practice. Some insight in my situation... 

Yes the home is well above 20k sqft at multiple levels; I also need to cover a swimming pool sitting 50 feet away from the home structure (position of the closest Sateliite). Note: I did state that WAPs are not closer than 50 ft apart "line of sight" with many walls of mixed construction to deal with concrete, plaster lath and sheetrock and studs. 

I use MetaGeek wifi analysis tool with channelizer software to view the radio spectrum frequencies and energies. My past roles in telecom does  include RF signal analysis for FM, satellite, microwave and mobile radios. (I was engineer for responsible for IBM’s microwave system spanning from Madison Ave in NYC, up the Hudson River, to Alpine, NJ, White Plains, NY, Poughkeepsie, Fishkill and Kingston NY. Corporate performance on this system back in the 80's was better than 1x10^9 BER for 18 years.  This was achieved by modifying conventional tower spacing, precise frequency bandwidth, antenna alignment and power output. Also did 39 earth stations in the US/Canada to bypass excessive carrier tariffs and mobile radio schema while at BlackBerry. 

In thjis estate, the closet neighbor is several miles away and the only radios in WiFi spectrum are those we generate in the home. 

As you know... overdriving the RF signals would result in issues that would 'appear'

 in the for of packet loss/retransmission, reduce window sizes and packet/bit errors. None of these were measured during test. Keep in mind that antenna design is important as signals travel radially (perpindicular) from the antenna. Thus in multi-level situations, RSS (relative signal strengths) drop off rather quickly.  {{ Just a trick of the trade... if you need to travel vertically where these is no wiring, take two WiFi devices and lay their antennas flat/horizontally) and you'll have a wireless link going up several floors!). 

The fundamental issue here is the Orbi Product not finding its own neighboring nodes on a completely wired multi-gigabit/10 gigabit L2 backhaul. There are no dependencies on wireless technology whatsoever! That's is a control plane problem (akin to how OSPF would be used in legacy networks to find neighbors and shortest paths across the interior network). NG has really messed up here with what appears to be shortcuts to put all thing wired and wireless into the same basket! The inability for the router to see its wired neighbors over the wired backhaul is totally inexcusable. RF interference matters can't be the root cause for the control plane not converging. Wired client physically attached to the LAN side should get to the main router NO MATTER WHAT via the physical  fabric because there is no need or consideration for 'in-home roaming/mesh' techniques.

But as a general practice, your point is very valid as a general practice. If one has spectral measurement gear, you have a chance to exploit the rules to optimize the actual environment. 

BTW.. your responses and comments (those I’ve read) were very much ON POINT. I can see you are a pro in this field and certainly have my respect.   …marcel

Sorry for the delay @furrye8...

I had an NG unmanaged 10G switch in at first. Seemed RBS' didn't like it. So ,I fell back to TP-Link TL-SX105 and TL-SX1008. I've had no problems in a dozen installs with the TP-Link products with ORBI in the past.

Sorry.. in rushing I failed to see your other questions:

• The house over these past 20 years has every Cat # from 5, 5e, 6, and 6a. I run the Orbi backhaul over Cat 6A.  STP is unnecessary, speaking as an "EE". Shielding is only useful if it is properly grounded. "Proper Grounding" however, is a simplex transmission problem! You ground shields at the differential input circuit, not the output circuit.  This is classic in the design of RF and instrumentation equipment. The goal is to short the non-common mode noise over the cable at the point just before the input. Common mode noise is dealt with by the twists in the pairs. The dif amp at the input cancels that noise out. If you look up special twisted pair cables you would see a shield over each directional pair so you ground at the appropriate end. Category cable was never this precise so you see just one shield over all conductors. The theory would have it, that if you wish to ground that cable as well as you could possibly do, it would be grounded in the absolute center of its length! HA HA. Reason being tis that you would only have non-common mode noise picked up over only HALF of the cables run. 
• Long convergence time to recognize nodes on the Orbi control plane goes well beyond wiring. All cables were validated (that's well beyond pin continuity checking) at 1 Gbps using 1 minute BERT testing. Just an FYI... you can run 10G on 30 feet of Cat 5e! it's about two things: (1) "puffs per foot" that is pico farads per foot and (2) whether the terminations at each end were done properly to avoid Near and Far End crosstalk. I've had the fortune/misfortune to design/review/test cables for huge IBM, AT&T and BlackBerry data centers (hundreds of thousands of cables). 
• Don't have floor plans but what most folks will dwell on are distances when the more important factor is construction composition. I've had a job where standing on each side of a single wall had ZERO WiFi signal when there was only 4 feet between router and access point. Problem was it was prior home owner was an engineer and built a Faraday cage - 6-in thick concrete wall with embedded wire mesh. But my problem is not about devices. It's about the router not seeing the satellites over WIRED backhaul, albeit 6 satellites. 
• Distance (independent of construction composition) AND if on the same floor would be around 30 feet. My radio goal is to have an RF signal around -40 to -50 dbm signal in the places where the home owner works, sleeps, sits. This is what I achieved with the system I installed. 
• You're right:  Disabling dynamic MAC is correct per Apple's recommendations - but this will not help the situation of the satellites connecting to and disconnecting from the main router. 
• I'm happy that you are able to get excellent wifi in your home. I started in commercial telecom in 1975 (telex & telegram). Worked my way to be responsible for IBM Global Network, later AT&T Global Managed services and then BlackBerry's infrastructure engineering. Today, I do smart home/networking in residential for the past 20 years. I retired my corporate engineering position 10 years ago where building data centers and transmission systems and dealing with Cisco, Avaya, Juniper, Lucent Technology and every carrier in 60 countries was my executive responsibility.
• Your points are very valid.

I do think however, this problem is an inner control plane problem coupled with a limited design for mesh implementation. The interior problem is what causes the known state of the RBS and RBR to not 'know' if they are up or down and what the best path to get between two points on their interior graph is. That coupled with a questionable implementation of mesh and the roaming that needs to happen. Adding 'management' to these is also a likely sore point in this design. They were in a max of 4 satellites since the introduction of ORBI while their competitors have already achieved as many as eight (ASUS). {By the way, it took ASUS over 3 years to figure it out, so this is not a trivial problem. "Mesh routing" was only introduced in mid 2010's for the commercial accounts. and much later for residential use.} 

@Furye38: thank you for your insightful posts. Yours, donawalt and OrbiTechDude's allowed me to figure out what I had to do... much appreciated. ... marcel   

Thank you OrbiTechDude. Then all this mishap was due to my own lack of observation... I learned the hard way.... thank you...marcel

Thanks... you had me chuckling as I imagine our having very similar experiences! Interestingly, Amazon had a 'deal' on the Orbi 97X's and limited you from buying more than one 2-RBS pack in a single purchase. So, when I first received the 3-pack and one 2-pack, I set up the RBR and 4 RBS's. Very stable results for the first 3 days. All wired client devices were correctly seen by the Orbi app. It was when I added the 5th RBS that all hell broke loose. Things that were stable before, were no longer stable. reboots, loss of RBS connectivity (per the web and app pages) were declared. From a basic network capability however, these 'not connected per the UI satellites were in fact on the wired network. I could ping them. Moreover, I took my 200 ft CAT6A extension cable roll and attached one of the RBS to it. I then ensured my iPhone was connected to 'that' RBS and ran near gigabit WiFi over the so-called "not connected" satellite (again according to the router, the Orbi app and web page). 

My focus was solely seeking stability in the Orbi RBS/RBR wired network without regard and consideration of any client device attachment or other wireless matters. When NG support told me that 9.13..;. would not help stability and that I should reduce to four satellites, I realized I was in serious trouble. 

Your point of balancing distance when using wireless backhaul is similar to what I've seen and done; it's actually the reason I do not mess around with wireless backhaul as a general practice. N.B. I do use MOCA 2.1 adapters in installs where it is impossible to run new Cat 6 or better cable. Though I've not tested it at this speed, they're said to support 2.5 gbps. My test gear allows me to test & validate/qualify up to 1 gbps.

I use an app written for windows platforms to assess the wireless neighborhood. Company is called Metageek that make a USB dongle and software called Channelizer. re: metageek.com  With this product you can find identify the types of sources occupying 2.4, 5 and 6 ghz spectrum. It data logs the RF energy so you can go back and analyze what is going on. With a directional antenna, you can actually 'hunt down' the sources that may be interfering with you. They provide characteristic profile RF energy curves that help you determine whether the sources are Bluetooth, zigbee, zwave,  nanny cam, pro-security cam, or other devices. My approach anecdotally speaking is to get a satellite to within 25 feet of the client's most frequented locations. I've found that the wireless throughput seriously drops off when distances go over 25 feet (some studies show that WiFi throughput seriously drops off at that distance); that's without any walls in the path. I end up adjusting transmit power get optimize performance and minimize interference; with all satellites using wired backhaul, the problem reduces to only being a client device problem and not a satellite to router wireless problem.     

Wow, 1862 construction! I'm sure you've had your WiFi challenges with the different materials used in your home over the century. Impressive. 

Were the drop off issues meaning a drop-off in throughput as WiFi distances increase or the client device randomly disconnects from the Wi-Fi. If it's the later, I was wondering whether you suspect MLO operation? My past experiences with 'bonding' or 'stripping' methods over parallel paths have been very erratic. It would be interesting to know if turning off MLO (or any other 'smart connect' function would improve the drop-off situation.   For decades we've had separate SSID's for each band (2.4, 5) can device drop-outs hadn't occurred. it wasn't until some bonding technique to leverage 2.4 & 5Ghz and now 2.4, 5 and 6ghz did device drop offs started happening for me. Two other related questions are (1) whether the drop offs are occurring based on 'age' of device (meaning older verses newer devices), and (2)  whether the drop offs are due by some 'power saving' mode in the device. 

Be well.. thanks for your interesting dialog...m