Dear all,
thank you for your responses. If nobody else beats me to it, I will try to write a Pandoc CSV writer in July, when I have a little bit more time.
My argument that CSV is actually a perfect fit for Markdown and a natural way to describe tables apparently didn't convince this list. In my view the current table implementations supported by Pandoc do their job, but don't look like a good fit for the Markdown philosophy of keeping things simple yet powerful. Maybe one reason why there is no standard Markdown syntax for tables. I know that there are various ways to accomplish CSV tables with Pandoc filters, etc., so it is not a practical limitation.
As far as using the ![] syntax to load other content besides images, that is a very interesting proposition. Given that the Github issue is from 2012, I am not sure how much consensus is on that topic.
Best,
Martin