What confuses me is that tx with .0001 txfee (about 40 sats per byte) are often delayed, why are lower fee tx taking priority?

My guess is old node software heuristics (?) that prioritize differently.

As for the 30%, I think early-to-mid 2017 probably. But by then it'll be 1.65mb effective due to segwit. Even if half the mining nodes use it, the bump will keep it ok. Hopefully segwit doesn't have some major clusterf*** in it. I have a bad feeling about it.

(end of nested replies, lol)