All great discussion, don’t have a clear solution on management fees, but i would like to chime in an bring some color from the inside process. I helped implement some of the systems and process @banteg mentioned above.
As we know a strategy doesn’t appear out of thin air, apart from the author, it needs experienced code reviewers (strategists) first that need to be paid somehow (this is extra work, not performance yield), it needs security reviewers (also paid or incentivized somehow), audits most likely eventually (also not free) and then infrastructure for alerting/adjustements/management, yes is needed, users ask for yield estimates, harvests alerts,etc, all of this needs code that are not smart contracts and somebody needs to build and support also not part of strategy code and doesn’t drop from the sky.
Once the strategy is up on chain, it really depends on the nature of the strategy what level of monitoring or management it needs, but the discussion on 2% management fee assumes for a simple LP to curve type strategy where the emissions are farmed, i would say that’s fair to ask for adjustment if the strategy can be somewhat this simple, but this is not 2020 anymore, most newer and higher grossing strategies are not simple.
Strategies that have high yield usually are not simple LP/staking strats, most are either leveraged, debt based that are riskier by nature and 10x more complex to code, review, audit and manage. The code can work in several good and bad conditions and we code and test for that, but the code is limited to what happens on chain on its boundary, blacks swans can happen. Even a trusted system/protocol like compound had an issue that affected several of our bread and butter strategies.
No AI or automated bot would have been able to decide to reduce our exposure on its own, not saying it cant be done, but current protocol version doesn’t work like that yet, doesn’t mean we stop trying to move to more automated yearn, im for that actually, but fact of the matter we need to keep a float to be able to keep innovating and current iteration of the protocol works this way and needs a fair amount off chain work and infra apart from just smart contracts.
Even Curve LP strats that at this point can be called simpler, need an active off chain tracking of stats to be able to predict the optimal voting % for gauges that would give the vault holders the best yield going forward, this is currently unfeasible to do onchain. Somebody has to do this data analysis, also doesnt fall from the sky.
Hope that helps bring some color to the discussion, i do agree we need to explore options and be flexible with fees to benefit users and holders, but we need to have the complete picture to make a correct adjustment that would not hamper our current processes.