Attempting to summarize:
- If attester yield gets too low, the ratio between issuance yield (y_i) and other kinds of yield - REV (y_v) + restaking (y_c) - becomes tilted more in favor of those other kinds of yield.
- If y_i is too low, the incentive to be an honest attester breaks down, and we may see clients caring less and less about this functionality. Clients may even stop attesting entirely, because this is a clean way to protect them from slashing.
- If we try to cap the total staked ETH by having an issuance curve that approaches zero or negative beyond some level, then this creates a risk, because if total staked ETH does approach those levels, incentives to attest break down.
- We could solve this problem by increasing non-participation penalties rather than participation rewards (see beacon chain spec here). But this would change another variable: the minimum portion of time required to be online to break even. Currently, this portion is ~0.5; tripling non-participation penalties (so, doubling the size of incentive to participate without changing rewards) would increase it to ~0.75.
- y_i is inherently near-zero variance, and y_v is inherently very high variance. Hence, reducing y_i would increase the percent of validator rewards that is highly variable.
- Recommend decreasing issuance moderately at high staking levels, by changing issuance curve from \frac{cF}{\sqrt{D}} to \frac{cF}{\sqrt{D}(1 + kD)}
Does this feel like a reasonable summary?