Is there evidence on the effects of extreme claims on moral circle expansion? 

I am relatively aware of the potential benefits (eg. talking about shrimp sentience might encourage people to start reflecting on the welfare of farm animals more broadly).
But it feels like there is also a potential failure mode with this current approach, where talking about extremely non-mainstream ideas such as stopping boiling vegetables or refraining from using antibacterial deodorant (as per the article) might make readers less likely to engage with similar arguments in the future.

Curious to hear if anyone has more insights on that!

Charity entrepreneurship has led a pilot project somewhat related to your point (2). There's more info in their writeup. 

One thing to keep in mind with PF doing "ineffective"grants is that they're sometimes limited by their statutes, eg. the Hans-Wilsdorf foundation, despite moving multiple hundreds of millions of $, can only do so in the canton of Geneva because the founder decided so when creating the org.