The problem with doing a client for this is lack of scale. Assuming there's 100k active users, 50% run iOS/Mac, probably < 10% of those would pay for an app. I'm thinking ~5k copies which sucks compared to the effort required.
@pth How about a "minimal effort" version, entirely based on Tweetbot, where you just change whatever is required backend-wise? The only entirely new thing would be a "choose instance" page. Starting off slow would make it worthwile even for only 5k copies at $10 each, I believe.