While I partly agree with your concerns, I think we need to put a bit of a perspective on that.
The bull market brought a lot of people, and ever since the bears have taken over it goes down. A lot of people jumping in back then lost money, and faith in crypto in general. The graph shows a steep drop with the last HF, as does the price chart. Engagement in generall seems to follow price more than anything.
Manual curation actually has gone up. How many people struggled to get over a dollar before the fork? Most who stuck around are better off now. Maybe not with recurring content like contests or automated statistics - except they have autovoters piling up on them, which I ranted about a few weeks ago getting a lot of backlash - but artists and writers and travelers generally have improved.
Even axeman isn't doing too bad. He can't expect to post a single photo 5+ times a day and make $10+ on each. Expectations have to be managed - $10 are 70 STEEM!
Also, not everyone busy with downvotes doesn't have time for curation any more. Look at @curangel - with its trail it seems to be responsible for 1% of the total votes (which I just realized looking at those charts, and it frightens me quite a bit...). We're far from having 1% of the active stake :D But with that project as an example I also can't confirm your issue with the curation rewards. We're doing pretty well, voting consistently on small posts.
I actually hoped that more people would use curation services like curangel with newsteem to spread their votes, but obviously that didn't happen to the extent that might have been desirable. The good thing about less votes being cast is that each one is worth more ;)
So yes - whales piling up on a few people are a huge problem, as that lowers rewards of everyone else. But it's never been much different. Rewards are lower than people are used to across the board. But were not seeing payouts in the hundreds for the top any more, as it was before (with similar or even lower prices). It's spreading slowly, but it's spreading.