The distinction between Good Node and Full Node is binary. Maybe it gets smoothed out over an entire network, but it could also give instability.
It would be better to replace Good Nodes and Full Nodes with a single value for % filled. I know the metrics on available space is the crux. (And part of another discussion)
Additionally, farmed coin (balances) of entire network doesn’t seem too complicated to know. Couldn’t there be a continuous pulse, a recursive request repeated between all neighbouring sections, where the total balance of each section is included, (and signed).
An approximate value for data storage available and used for each section, could be included there as well.
This could then be used in the farming algorithm.
Scaling frequency of the pulse would be one matter, also how deeply recursive it needs to be. Probably not very deep at all, before we have e 100% of sections having the data very close to network average.
Going only by the single section’s data will give that a few sections on either end of the normal distribution generate extreme values.
Because even though data will be distributed randomly in the network, it will not be evenly distributed, and so it does not prevent huge differences between some sections occurring.
Even if the neighbors metrics sharing is not something that is desired, I would like to see some analysis on what consequences the occurrence of sections with huge differences in generated values give, and how it is planned to be handled. (Or shown how it would not occur at all in the first place).