I think the whole referral implementation is flawed and we're trying to tweak the wrong things.
I think the problem is we're trying to build a public mandatory referral system and tax, when it should be left to private business, private interest to offer these thing if they are useful.
I am interested in exploring options that completely reduce the friction for users, reducing the fees, reducing the tax. Where you've got a wide open ridiculously efficient platform with little to no cost at the core and offering innumerable optional features with many incentives to participate or utilize, like a variety of optional referral programs, built into privately created assets like FBAs or smart contracts, or just webservices etc. that might be completely separate from bitshares core.
Could we not have many different (optional)referral programs built on top of bitshares? User can choose the one they like, or none at all.
Does the referral program need to be centralized as a core universal unavoidable feature/tax?
Couldn't businesses develop their own referral programs, incentives for using their wallet, their service, etc that don't rely on a core design mechanism that has a significant measurable effect on every user and every use case?
If a referral system has merit people will voluntarily use it. I think business should design their own programs with their own source of income from their own services and not rely on a core tax on bitshares system wide. It's a burden and inefficient.
A business might create a bitshares smart contract or a new type of transaction or feature or process or game or service etc that they design to collect a tax/fee that feeds their particular referral program. But they would be subject to competition from competing services or competing referral programs and no one would be forced to fund them or participate in them while using the bitshares platform.
I don't like that this referral tax is unavoidable. I think referral programs are a great idea but they should be built on top of, and in addition to bitshares, maybe not baked in the core.
Anyway, that is the direction of my thoughts lately. I could be wrong.