BIP101 is the only proposition which is more than a concept and has actual code.
I think participants will re-evaluate if we bump into the limit and it causes some inconvenience.
Miners cannot afford unhappy users, their business would be ruined quite quickly.
Wrong. This is a simplified version of a possible scenario based on (nearly) zero evidence. To clear up: 1) Miners don't really care about you; 2) Businesses would not be ruined because of a backlog of transactions, certainly not quickly; 3) People who want their transactions processed sooner need only to include a higher fee; it is that simple. Anyone saying that the fees are going to become $1+ instantly is spreading nonsense. BIP101 is one of the worst possible "solutions" that I've read about. It comes close to the 'no limit at all' idea.
Tl;dr: Nobody can predict the future, so stop trying to predict the block size.
I'm not saying that I can predict anything, only my opinion. (mind the I think part)
As for your opinion that miners do not give a shit about users is obviously wrong. Without an userbase/satisfied participants they have no bitcoin mining business.
Too big of a backlog certainly affects transacting entities negatively. Saying otherwise is BS. I think in a lot of scenarios it does matter how fast the tx gets included in a block.
Higher fees will lead to people searching a cheaper alternative. As for the amount of fees, it depends on the backlog and number of transacting parties. I think everyone shared goal is to have as many participants as possible, that will lead to more utility, more robust and stronger network and markets.
I think BIP101 is the best solution as it does not give more power to miners than necessary like bip100 (that would open up an other can of worms). I would love to see actual tests (testnet?) with unlimited blocks .
Mind that I'm open to alternatives, but the keeping of a crippling limit.