Ultimately I want access to the Twitter Ads API, at which point I'll switch from favorites to serving promoted tweets/accounts. I think I've innovated on the Ads UX and the form which the impression takes is secondary. This is a surrogate impression. It has allowed me to build an Ads product and an Ads customer base in anticipation of Ads API access.
What you've done is move ads from the place where they're supposed to be (promoted tweets) to the place where they're not supposed to be (your interactions). Innovative spam is still spam.
There are versions of this approach all over Instagram too - I 'll wake up to a large number of favorites, only to find out that most of them were bots. (Although, with the OP's approach, the accounts that are favoriting this are real accounts, so I wouldn't be able to tell easily by going to their profile)
If I get a favorites notification, I want to go go their profile and find out if they share similar tastes. If it's just an automated system, they have ruined this for me by lowering the signal-to-noise ratio.
Regardless of how innovative this is, Twitter shouldn't get any flak if they decide to shut him down.
The distinction for me isn't between marketing/non-marketing but between humans and robots. If a human representing some business favorites my tweet because they feel that it's worth favoriting, that's great. In fact, it makes me feel good that some human at an organization is reading and caring about what I have to say. Wonderful, I write things so people can read them.
Robots just going out and marking things as favorites just to get attention and followers? That's the kind of spam actively undermines the value of the example above.
Yes, your customer list may be useful to Twitter - as a list of accounts to suspend for violating their ToS, devaluing favorites, and hurting any legitimacy Twitter had.
You shouldn't have existed (or been allowed to) in the fist place.