Businesses have spent several years building based on the current reward model.

The question is if businesses were built because or in spite of the current reward model. :)

I got the sense that they want to "punish" self upvotes but not all self upvotes are created equal.

A convergent linear rewards curve would still allow everybody to upvote himself - it just would be a little bit less beneficial.
I also upvote myself and good friends, but at the same time I would see an incentive for everybody to spread his upvotes on as many different accounts as possible as step into the right direction. Too many new user earn nearly nothing, even if they deliver very good content, while many more established ones mutually upvote themselves and get the biggest part of the cake (apart from the bid bots which I consider as harmful, too). You may reply that the ones who have invested money also deserve the biggest rewards ... However, I would happily prefer to get a smaller part of the cake if then at the same time the cake was getting bigger (the STEEM value higher) again, instead of smaller and smaller ...
And that's the crux in my opinion: the more satisfied users are on STEEM the worthier it is for potential big investors to use STEEM (and not any other platform) for advertisement purposes, to offer their products or utilize it as communication channel, which longterm would make the STEEM price raise again.
Apart from that we should not forget that the high value of Facebook is due to the huge amount of its users. The value of a (social) network is measured among others by the number of its users. I would be more than happy about a somewhat weaker self-upvote if at the same time my STEEM would have a higher value again.

It's completely OK, if you disagree with the above, but I felt I had to contribute my opinion here. In case you like to read more about my ideas to improve STEEM, you may read my last article.