I think I am getting somewhere with the Flashee wear levelling library. I found that data blocks between 20 and 60 bytes made no difference to the problem which occurred when the Core was powered off and on. It appears stable on all my three Cores when just applying a hard reset with the button. I then thought about my switch and applied a reverse diode (preventing inductive kick?) across the power supply and then a large capacitor to delay power down. Neither of these made any difference.
The latest Deep Update patch that I applied to one of my ‘black’ Cores caused my original Cloud enabled code to fail. However when I loaded the local TCP/IP code into it with Could disabled the wear levelling problem had disappeared.
So far as Cloud storage is concerned I expect there is room for both local and Cloud storage. Once I am confident that the Cores are stable enough in Cloud or local connection mode I will test the market and determine what the likely demand will be.