I have discovered two very important problems and would like to know if anyone else has found them:

Using V1.2 of the PCB there is an error in the printing of the text. 'SEL' and 'RST' are the wrong way round! This is easily confirmed with a continuity check.

Second problem is deeper. The RFM95 module does not work with CS permanently held low because it seems to use the falling edge to set up a package exchange. It has 3 modes including a burst mode for reading or writing consecutive addresses and I think this may be why holding CS low does not work.

I solved this by using D1 for CS - I did not use pull ups on D1 & D2 because I am not using I2C. A short wire from the pad for the I2C chip to the pad marked RST (actually it is SEL) now allows the RFM95 to work correctly and I can use D0 to wake from deepSleep.

If anyone else is working on deepSleep this may be useful information and I would be very interested in your progress. I have more details if anyone would like to contact me.