I think I have fixed the issue. I changed to the cap on the ethernet gateway from 47uF to 1uF. I have power cycled the gateway 4 times now and all the nodes came in. Looks like less is more! For whatever reason the breakout boards obviously don't work well for me.
I will keep testing for a few days.

@mfalkvidd No I only use one at any time.
I started with serial and then built the ethernet gateway. To eliminate the ethernet gateway hardware I just reverted back to serial to test.

@yveaux I have a serial gateway built on Arduino Uno connected to Pi. This is now just spare as I have W5100 / arduino nano ethernet gateway. Nodes are mostly 3.3V pro minis. One is 5V nano. All arduinos are clones.
Radios on nodes are NRF24L01+. On gateways I can use NRF24L01+ or amplified version (doesn't make a difference to this behaviour).

As an example I switched off the gateway 2 hours ago. 2 nodes are came in straight away, the other 3 have no communication. Chances are the same 2 will connect again if I power cycle the gateway but the nodes that work straight away changes randomly. It seems like the gateway does not want to acknowledge certain nodes.

Behaviour is the same if I revert to stock motion sketch.

The parser says that no radio ACK has been received when the node tries to ping the gateway.
I have today built a replacement radio breakout board from scratch. The problem is still the same.
Some nodes connect, some randomly do not.
I have replaced every piece of gateway hardware, so this my be something within the Mysensors code.
Any suggestions please?

@mfalkvidd Thanks for that.

F11 showed that the page was failing to resolve cdnjs.cloudflare.com
I run a pihole so just needed to add to my whitelist.

Fixed. Thanks.

Now back to the node/gateway communication problem!

@flyingdomotic Ok So I have modified \Arduino\libraries\MySensors\hal\transport\RF24\driver\RF24.cpp and reuploaded my sketches to the gateway and a couple of nodes.

Unfortunately one of those nodes is still reluctant to connect. You can see the tx LED coming on for a long time while it tries.

There must be something in the underlying code not quite right?

@electrik Thanks for this.
I am on Windows 7. I will try another PC. EDIT HAve tried another windows 7 machine No output on IE Firefox Chrome or Opera?
As far as i know the breakout boards I am using have capacitors included on the. So I just apply a good 5V and plug in a radio.

@FlyingDomotic Thanks. So would I put that code at the top of the sketch for both node and gateway? Before the loop?

If I restart my gateway (tested with serial and ethernet gateways) some nodes connect straight away, but others can take minutes or hours to connect.
I am thinking this is the root of all my connection issues.
NRF24 modules of various types tried.
One day certain nodes will be problematic, another day other nodes. I have them all very near to the gateway.
I use breakout boards for the NRFs on the gateways without any extra capacitors/ Could that be it?
I cant get the log parser to work in any browser so not sure how to decode.
The node produces the following debug when it cannot connect:

4130 TSM:ID
4132 TSM:ID:OK
4134 TSM:UPL
4171 !TSF:MSG:SEND,2-2-0-0,s=255,c=3,t=24,pt=1,l=1,sg=0,ft=0,st=NACK:1
6180 TSM:UPL
6217 !TSF:MSG:SEND,2-2-0-0,s=255,c=3,t=24,pt=1,l=1,sg=0,ft=1,st=NACK:1
8224 TSM:UPL
8261 !TSF:MSG:SEND,2-2-0-0,s=255,c=3,t=24,pt=1,l=1,sg=0,ft=2,st=NACK:1
10268 TSM:UPL
10305 !TSF:MSG:SEND,2-2-0-0,s=255,c=3,t=24,pt=1,l=1,sg=0,ft=3,st=NACK:1
12314 !TSM:UPL:FAIL```

@flopp As far as I know everything is running on 2.3.1
I now have all my nodes very close to the gateway. Will keep testing for a few days.

@sergio-rius Each node sends a battery voltage every 2 minutes. If the PIR is tripped then it wakes from sleep and sends PIR tripped and also the battery voltage.
I currently have Domoticz set to send me an email if it does not receive the battery voltage. I have tried MYScontroller but not sure how I would monitor all those messages over a few days? Does Domoticz have issues?

I have currently taken the repeater out of the equation and replaced the NRF24L01+PA+NLA on the gateway with a NRF24L01+. So far only 2 errors in 24 hours from the furthest away sensor. Have moved it closer for testing.

Update - after 2 hours I got the first failed message. I can understand that nodes may try to send at the same time but there must be a way to keep retrying?