Had a quick google, couldn't find a corpus of spam blog posts. Surely somebody has one?

I hear someone just released the full corpus of Reddit comments...

Also use a chatbot script to randomly reply to the user's comments.

Part of the ethical construct of the current form is that there is no trickery. The user sees the real, unadulterated data.

What about Markov chain reconstructions of their own comments?

There is something similar to that here: https://www.reddit.com/r/SubredditSimulator/

The only difference is that you cannot comment.

My favorite place on Reddit. It's a goldmine of ridiculousness.

Markov chains need lots of training data to be useful. If you feed it just one comment, it pretty much reproduces that exact comment, maybe switching around or repeating a phrase somewhat.

Given the target commenter user base, I don't see that being an issue. ;)

That is true. It would also betray the name echochamber. :)