RyanF

Further to my earlier post which was released in error......
A few months ago I bought 4 x 3.2 volt 100 Ahr cells (new) from a company called Cinch-IT in the Cape. 12 volt BMS (excluding temperature sensor) was bought on a special from a supplier advertising on BidorBuy.
I also purchased 4 x 3.2 volt 50 Ahr cells (2nd life) from Electromann as I had already purchased a 12 volt 55 Ahr BMS (with temperature sensor) from them a year or so earlier. Must say that these cells looked the worse for wear but so far have behaved faultlessly. 
Aluminium flat bar were used for the links/busbars. A2 stainless steel hardware was used throughout. Purchased a battery box ex. Electromann for the 100 Ahr pack and used an off-the-shelf ABS electrical box which I modified to suit the 55 Ahr pack.
Various blogs stated that all the cells should be brought up to the same state of charge using numerous methods. I chose to charge up each cell separately (bit laborious) with a dedicated single cell charger (bought cheap ex. Battery Experts) prior to coupling everything up.
Was it worth it? Depends primarily on the price and quality of the cells. Also, good quality BMS's are not cheap. In addition, don't expect much in the way of component guarantees if it is a DIY project. If you make a stuff-up you could be left with an expensive paperweight at best - or potentially cause real harm to oneself. Safety first! 
All in all I saved around R3000 on the 100 Ahr build and close to R2000 on the 50 Ahr build (excluding my time) when comparing them to middle of the range batteries. Obviously not talking Blue Nova prices here!      
       

Sorry for the long delay @conradl
Below is a simplified diagram of what I have. Hope this helps.

Or when you want it less complicated you can give Hubitat a try. Very nice hub!

Sorry for the long delay @conradl
Below is a simplified diagram of what I have. Hope this helps.

@AntonySo after trying two depths of heat trap loops (150mm first, then 600mm on both inlet and outlet), they really don't seem to work much (if at all). After the heat traps, the next iteration was to install a flap type NRV (horizontally positioned so the flap would close under reverse flow conditions) on the pipe which flows from the panel to the hot side of the geyser. I expected that ALL of the thermo-syphoning was going out the top of the geyser and the flap-type NRV would sort the problem. To my surprise the thermo-syphoning continued and when I had a close look at the data it was clear that it was going out of the cold side of the geyser up to the panel during the night (by this stage I had installed a temperature probe on both the inlet and outlet of the geyser). Some posts suggested a spring type NRV on the cold side, but I was not convinced because during the day the water has to flow in the same direction as the thermo-syphoning would go at night. But to my surprise it stopped the thermo-syphoning - I guess the spring is just strong enough to stop the little pressure 'bursts' which characterize thermo-syphoning - but not strong enough to prevent the pump flow during the day. For the fist time the panel temperature dropped well below the geyser temp for the whole night!
So my experience is that you need two NRVs - one on each line, but the cold side one has to be the spring type. Probably worth trying before re-piping 😰. Just be careful to get the NRV directions correct so you don't end up with a closed loop 🥵
FYI - Below is a graph of the data when the thermo-syphoning was happening out of the 'cold' side of the geyser. The blue line is panel temp, the grey line is geyser hot side temp and the orange line is geyser cold side temp. The more regular 'up-and-down' of the orange line from around 23:00 is the small bursts of pressure that thermo-syphoning produces as it forces the hot water along and it is replaced by cold water. Overall you can see that the panel is hotter than the geyser during the night and they all have the same downward trend indicating the heat is being bled out of the panel.

I've seen recommendations of 6 inches minimum (Type L copper) vertical loop height within the first horizontally 6 inches of the cold water inlet and hot water outlet.