Question inspired by the responses to this: does anyone have a good fully loaded cost breakdown between running an L2 w/ centralized sequencer vs PoA L1? I would have to imagine L2 is cheaper (just by virtue of not having to share tokens with other vals) but what am I missing?