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.
Emphasis mine.
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.
2) There is one little detail here, removal of the limit on block size makes the orphaning a non-issue as it opens the door for targeting higher bandwidth layers (those fiber optics under the ocean are waiting

) where profit-driven miners would be incentivized to move sooner than later. They have little incentives to do that at the moment as the costs won't justify the effect with the current 1MB limit in place.
There are two costs that play against each other with Bitcoin.
1) First is the cost to transact on a single unified ledger that comes from the limit on block size (which incurs tx fees).
2) Second is cost to run a full validator of the protocol rules and transactions from home (ever growing blockchain).
The more transactions are allowed in the block the cheaper they become, but the costs of validating get higher as a result (and vice versa).
As of right now Bitcoin gives us two "keys" for our monetary sovereignty.
1) First is the "private key" from our coins that we can generate in a permissionless way and then use to receive and send money.
2) Second is the "key" to validate the fact that money we hold and use operate as intended by having permissionless access to blockchain.
The fees we pay in the network is the price to hold the second key, don't lose it!
Well put.