We should be able to test this empirically: if the theory is true, then, for example, we should never have a sustained period in the future where the total fees collected by miners is significantly greater than the aggregate losses due to orphaning.
1) We definitely should. We must let the pressure build up and see if the network effect alone can hold it. Maybe even let the pressure leave for a while.
The question is whether or not
we have a choice. My prediction is that there will never be a sustained period in the future where aggregate fees are significantly greater than aggregate losses due to orphaning. The protocol will fork (or demand will leak somewhere else) before this happens.
Let's say "sustained" = more than 6 months and "significantly greater" = more than double.