Thoughts?

I'm indifferent.
Also, I'm not quite sure what the point of this is.

Our stated downvote policy is "downvote what you want less of."
If we don't want to see "meta" discussions on LW, it follows that we should downvote them.

If enough people follow this policy and share this judgment (net), meta discussions will quickly be downvoted to invisibility. If not, then they won't.

Since meta discussions aren't quickly being downvoted to invisibility, I infer that not many people share this judgment and/or that not many of the people who do share it follow existing stated policy. Either way, it would surprise me if stating a new policy changed much of anything, though I don't object to you trying the experiment.

That said, committing to following existing downvote policy (either instead of or in addition to) might do more good.

Along the same lines, encouraging downvoting... for example, by periodically praising in public users who have used a lot of downvotes, if there's some way of looking that up.. might be a good idea.

I also suspect that this whole concern with whether a discussion is "meta" or not is serving as an indirect measure of discussion quality. I distrust optimizing for indirect measures generally, as it tends to decouple the correlation more than it tends to optimize for what I actually care about.
That said, I admit that "tends to" != "always" and this might be one of the exceptions. Again, I don't object to you trying the experiment, though I expect the results to be indeterminate.