Maybe I'm missing something, please correct me if I'm wrong.
In the event of an empty mempool (and with a 0 btc subsidy), for a miner to obtain a profit would be necesary to wait until enough new fee paying transactions enter into the mempool, otherwise keep mining at a loss.
Isn't this the same as saying
"In order for a transaction to be added to the blockchain, a sufficient fee must be provided"
In the event of an empty mempool (and with a 0 btc subsidy), for a miner to obtain a profit would be necesary to wait until enough new fee paying transactions enter into the mempool, otherwise keep mining at a loss.
Isn't this the same as saying
"In order for a transaction to be added to the blockchain, a sufficient fee must be provided"
It's a matter of whether we're talking about the fee per transactions, or the total fees that a miner can claim when he builds a block. In the case where R->0, you could post a TX with a generous fee1 but it still won't make sense for a miner to attempt to mine a block just for you. You'd need to wait until others have broadcast similarly fee paying transactions, in order to push the block size into the "profitable" zone in Mengerian's graph:
1There would be some fee that you could pay to get your TX mined, but it would be vastly more costly than waiting for other fee-paying transactions to begin re-filling the miners' mempools.
The above discussion relates to Bitcoin in a relatively mature state, well up the S-curve. Even a negligible subsidy changes the dynamic from a zero subsidy, and the $ value of BTC might be so high that incentive to mine an empty block remains for several decades. Consider the current situation with low oil prices, many producers remain pumping at a loss while they wait for an upturn because selling at any reasonable price mitigates their fixed cost overheads.
Further, I don't think that the mempool will be emptied as standard, because it implies near 100% synchronization and there will be differences at the margins: new tx arriving and old tx expiring. One of the beauties of IBLT is that it will always allow for some out-of-band tx, which mitigates the risk of the centralised censorship inherent in "all nodes agreeing all unconfirmed tx" in advance of selection for block templates.