For CS tags there was a recent discussion here (to which, or rather the subsequent events, I implictly referred to in my comment when retagging your question) http://tea.mathoverflow.net/discussion/1380/arxiv-tags-was-cs-arxiv-tags/

So, on a practical level, recently all the existing CS arxiv tags got removed (or rather renamed to avoid the prefix), it thus seemed a bad idea to now recreate one cs.db in particular as there is an existing one named database-theory (which should cover about the same thing and is IMO more immediate to understand).

I did not participate in that other discussion. But my opinion was well-reflected in the discussion in that I find it useful to have some kind of top-level tags and the math arxiv ones happen to be a good fit (well-established, reasonable number [say MSC feels like to many], also perhaps that they are externally fixed has some upside).

Everything else seems first too complicated, and second I fail to see the point of having the prefix (for the math one the prefix signals the top-level nature). And the tag system seems already too complicated. Some people noticeably do not understand the prefixes already (creating prefixed tags that are not an arxiv category, ommiting the prefix and all kinds of things).

Let us keep things simple. More complicated things might look good in theory but will just not work in practise.