I mean dantheman has ~3mn $ in SP and his upvote is like 140$.
If curation rewards weighted exponentially more the earlier you vote as I heard they are (and especially if they are 3 - 5 times author rewards as claimed by @james-show in the comments here), then @dantheman's $140 in author rewards could be good for perhaps an order-of-magnitude higher in curation rewards. I know you were thinking about thresholding author rewards, but if curation rewards are more lucrative for pacts, then you'd need threshold them also.
Also I don't believe @dantheman's upvote is only $140. That is probably because his voting power was diluted by numerous upvotes. I've observed @smooth upvote my post and it jumps several $100s or more when he hasn't been too busy voting recently. @ned upvoted my first blog after my #introduceyourself post and I think it was like instant ~$1000 boost.
Even if you threshold at $500, it won't solve the problem. You will still have a minority of the posts at $500 and the vast majority at a few dollars. The top tail of the power-law distribution is very steep, so not many posts above $500 any way.
You'll need to bring the threshold down to much less than $100 per blog post to really spread out the rewards widely and then you will open the vulnerability to pacts.
Remember my calculation from our discussion at BCT:
Do the math. Rewards are 7.75% of market cap yearly with 50,000 signed up users, so 0.0775 × $200m ÷ (12 × 50000) = $26 monthly per user. But the rewards are distributed non-linearly roughly in a power-law distribution where the upper 20% get 85% of the rewards, thus 80% of the users will get 0.15 × 0.0775 × $200m ÷ (12 × 40000) = $5 monthly. If you argue that actual usership is 1/10th of that, then multiple those figures by 10... Social networks are typically valued at around $100 per user, if we are looking at the stable case down the line. Even if we up that to say $1000 with expectation of acceleration and more ecommerce than typical social networks ad funded models, it is still only $77.5 per year per user without factoring in the power-law distribution effect.
Also I assume naturally income and blog quality is going to power-law distributed as well in any case, so we won't be able to widely distribute rewards in any case. The only way to widely distribute money is to make a few people very rich and the 85% get the crumbs. This is unfortunately the natural power-law distribution of wealth that always exists in every instance of humanity. I guess we can argue the distribution in Steem is even more concentrated (worse) than a natural power-law distribution. So maybe you are advocating thresholds to try to get closer to a natural power-law case, but the problem is the vulnerability comes into play and we are starting from a worse than power-law distribution due to the way the Steem token was launched. So afaics we'd have to compensate with a very low threshold.