Well, i don't know if it is a timing issue or a hardware issue, but increasing the RFM95_POWERUP_DELAY_MS from 10 to 100ms solved the issue.
Now testing range improvement over rfm69h, will report back.
@tekka , Great work !

@NeverDie
Excelitas PYD1588/1598 are the best I've used, running now for years without a single false alert.
Fully integrated, 1.8V and only 3uA

Pcbway do produce better finished pcb, but the big advantage of jlc is their SMT service

You should define MY_SIGNAL_REPORT_ENABLED

@nagelc
I kept all setting to default except the maintain window =8s
So the SERIN reg is: 0x0031ED10

@NeverDie
With the fresnel and the wide angle 1598 it's ~ 7m

Pcbway do produce better finished pcb, but the big advantage of jlc is their SMT service

Sck = 15
Miso = 14
Mosi = 16
You choose SS and IRQ from available gpio

I'm using the sleepRemainingMs function for that and it works if accuracy is not important

They are programmed with the platform that you are using, assuming it support them.

@Anticimex more important is to set different network id or you'll have lot of collisions and lost packets

99% of the nacks are radio/power issues
You may recheck them carefully

You should define MY_SIGNAL_REPORT_ENABLED

@nagelc
I kept all setting to default except the maintain window =8s
So the SERIN reg is: 0x0031ED10

@NeverDie
With the fresnel and the wide angle 1598 it's ~ 7m

@NeverDie
Excelitas PYD1588/1598 are the best I've used, running now for years without a single false alert.
Fully integrated, 1.8V and only 3uA

If you are using pubsubclient, then

bool PubSubClient::connected()

should do the trick