I will have to think if an improved confirmation condition robust to multiple double spends could be thought up.
It is easy to track the total spend on transaction fees of descendants of the original transaction. A confirmation test of equivalent robustness to the Bitcoin 6 confirmation tests would wait until the total spend on transaction fees of descendants was higher than the total expenditure of Bitcoin miners in mining 6 blocks at present (including electricity, premises rent and amortized costs of equipment). This is a large number, but so too are total transaction fees if the coin is actually being used.
As you increase confirmation time, you further inconvenience all users, while the attacker only has to find more victims.
Also, you assume total transaction fees is something large when the coin is actively used, but I can't see (or I overlooked it) how you incentivize each individual user to pay higher fees.