I voted RAC but meant to vote MAG.
Here is why I think MAG is better. Over time the RAC is going to be very high (my prediction is 20M RAC by the end of the year, if not sooner). So if the decision is made to go on RAC it brings up the following conundrum. Right now at 1.5M RAC, lets say we choose 10BBP/RAC. While at 1.5M RAC, 10K RAC makes roughly 10K BBP/day. So right now you'd need to stake (at 10/RAC) 100K BBP to make 10K per day. The same machine at years end and 20M RAC would only make about 600BBP/day but still need a 100K stake. So either the BBP/RAC is so low that people running thousands of dollars worth of VPS per month that there is no sacrifice needed to mine the coin, or it is at a reasonable number now but at years end will be entirely predatory to new users who have no real chance of making anywhere close to their stake in a year, let alone weeks.
Whearas going off MAG will be reasonably equitable. The reward decreases over time (about 18%/year) and the RAC will likely be more evenly distributed (no one will likely have 10% by the end of the year).
As far as the sub 1000 RAC, I think just saying an account with one ARM based is exempt is enough, otherwise people will divide their accounts and cheat the system.
The only other system I see is far more complex and would likely confuse users. But the most fair system should be something like, you need 100x the BBP of your daily reward staked. That sort of system scales well and is equitable.
I don't think ARM is enough, not many people will want to use their phones to compute. Plus there is a huge iOS market that cannot run BOINC and the majority Windows novice computer users. In a couple of weeks or months 100,000 RAC might not give you that much reward. I'm not sure someone will want to create 100 VMs just to take advantage of the system. We can however change to perhaps 500 RAC, my Atom processor get 1000 RAC but that's running 24/7.