This is a great list, and is sure to start a well-needed discussion in the community.
The highest priority, IMO, should be the optimization of the 2,312 
/day distributed to each market, since small changes can have huge rampifications for how COMP is distributed, held, and used, and this currently represents almost the entirety of the distribution.
This is achieved by:
- Separating the compSpeed into separate supply & borrow incentives, work which is almost complete (any / all additional eyes, and auditing welcome; as a community we should bring this work to the finish line)
- Modifying the COMP distribution in ways which put more tokens into the hands of users; this likely means tying COMP more directly to “natural” use-cases, and less to “farmed” use-cases; collateral assets having COMP rewards only on the supply-side, and borrowing assets (stablecoins) configured with reserveFactors to make “folding” the asset un-economical (no “negative spreads”), but attractive to normal use. If done correctly, more COMP can get into the hands of users looking to participate in governance.
In terms of additional approaches / programs, as a guiding principle:
Expanding the grants team to encompass a full-time team has large upside–to encourage, and coordinate improvements to the protocol that are sustainable (long-term). Protocol improvements – whether on market selection, parameter selection, rate models, new features, and integrations are all results that compound. This is a great direction to explore, especially with the benefit of hindsight after months of operating the grants committee.
Another direction that the community can look, would be increasing the reserves of the protocol directly; figuring out a way to reward users that directly add reserves to the protocol, in markets that have the lowest ratios of reserves:borrowing. This is an alternative to providing COMP to farming users, while capturing most of the value for the protocol (and all users). This could increase safety & liquidity, while testing a new distribution model. It could be tied to vesting, etc.
Rewards for voting on proposals are interesting, but could create perverse behavior (e.g. making hundreds of “dummy” proposals, simply to capture more COMP). If this path is explored, it should be a set quantity of COMP per month, regardless of how many proposals there are, multiplied by participation rate. Now that there is an “abstain” option, it could bring participation close to 100%. This could even include a “quadratic” type reward, where smaller addresses earn a larger relative reward.