Whilst I do agree that consensus is not explicitly defined, as Bitcoin works better in practice than it does in theory [1], I disagree with the remainder of your statements. Redefining consensus on the go is exactly what the BU folk are doing, just in a slightly different fashion. Who are you to re-define consensus? Nobody.
As I said, defining a non-existent phenomenon has no meaning. I'm just using a word to designate a phenomenon. Like melting. Melting is an observable phenomenon ; if you use the word "melting" for some theoretical construct that doesn't happen, that's meaningless, and even if I'm the only guy using the word to indicate the actually happening phenomenon, I'm the only one using it correctly.
But I'm not even disagreeing with what others say. Of course, consensus is "that what most of the participants agree upon". Sure. But that is not what is interesting about distributed trustless consensus. What is interesting about it, is that there is a dynamical description that explains the immutability aspect of this "agreement", namely the fact that many diverse agents want to impose changes that suit them, and are unable to come to an agreement over that, so the old rules prevail.
It is amazing that people are so stubborn that they are discussing over how this is wrong, while they have exactly the example under their noses. This is like Galilei showing the 4 moons of Jupiter with his telescope to the representatives of the church, who refused to see them.
Where does immutability come from in distributed trustless environments, and how is this kind of consensus different from the "computer science consensus" ? Contrary to "computer science consensus", there is no "neutrality" in the nodes. In the computer science consensus, there are "honest nodes" trying to establish a pre-defined protocol, and there may be some faulty or malicious. But most are presumed to "try to be honest". This is different in a trustless environment of course. All nodes try to scam all others. If I use bitcoin, and I pay someone, of course, I would like to reverse my transaction after a while. Once I get the goods or services I paid for, of course I would like to "erase" my transaction from the block chain, to be able to double spend. All bitcoin participants are supposed to want to double spend, to modify transaction history, to make fake transactions in their name, and to get fake mined bitcoins. Nobody is to be trusted, and nobody is "trying to be honest". Everyone tries to scam everyone. That's the basic idea of a trustless environment.
However, in as much as I try to erase my transaction to you from the block chain, you want to keep it of course. I could collude with some others to redo the block containing my transactions, but I won't find enough people agreeing on that. I would like to make a false block giving me the coinbase reward, but others will not agree. So, no matter how much I would like to scam others, I won't be able to, because what I want to do in my advantage, is not in the advantage of others, and we can't agree over that. So we end up agreeing over the "honest" rules. Not because I'm an honest node (I'm not), not because you are a honest node (you're not), but because there's nothing else we can agree upon. You do not agree upon my scam, I do not agree upon your scam, and so we have no choice but to agree upon the honest blocks. That's why there is never agreement on redoing the block chain, and why it becomes immutable. I cannot make miners agree upon my scam. You cannot make miners agree upon your scam. So we end up, even if we both want to, not to scam one another. But nobody is trying to be honest. We are CONSTRAINED to be "honest" (that is, follow the original rules). We are FORCED to agree upon the honest history. That forced agreement is the emerging consensus, and is immutability.
This differs from the "computer science" consensus where honest nodes have no "gains" and do not try to scam other nodes, but just try to find the "right" data. In computer science, one has "distributed consensus". But not trustless (that is, everyone has an advantage in trying to be dishonest, but there is no collusion possible over this).
There is no "preference" to have. The dynamics is what it will be. Your question is like "do you prefer a high or a low oil market price ?". There will be an oil market price, emerging from offer and demand. There's not to be decided "what it should be".
Your analogy is completely flawed. You talk about the market, you talk about the economy, you talk about consensus, therefore my question is valid. If you can't answer it, then you need to start asking yourself some questions.
My analogy is not so bad. In a market too, everyone tries to get his advantage. If you sell apples, I would like to buy them for an extremely low price. You would like to sell them for an extremely high price. None of us can obtain what one wants (and scam the other one). So we agree upon a price none of us actually wanted, but is the "honest" one.
We see that the immutability doesn't work, when the parameter is entirely neutral, and nobody can use that parameter to obtain an advantage over someone else. This is why the normal word "consensus" (common agreement) is applicable for all those changes that do not imply any advantage of one node over another one. And this is why early modifications to the protocol were possible, as they were not subject to the "I try to scam you, you try to scam me, but the only thing we can agree upon, is to remain, against our wishes, honest, that is, follow the established rules".
This is why immutability didn't stop people change the block size when this didn't affect anyone. And why NOW, as it is affecting people and giving them advantage or disadvantage (when it becomes an economic parameter) it becomes immutable.
I'm writing this as someone who studies the dynamical phenomena of this kind of system. Not as someone who has any interest in it. I find the phenomena happening in bitcoin, as dynamical phenomena of systems, pretty fascinating, and I'm just sharing my observations with you, especially in these times when those dynamical phenomena have "unexpected" (to those who don't understand them) effects.
I'm telling you, the block size will not change, and segwit will not happen if my understanding of the immutability dynamics is correct. The only thing I see eventually happening (but far from sure), is a hard fork with two coins emerging when the tension is too large.
And honestly, the facts are in my favour. If it could have been changed, it would have been, since a long time.
Now, you go and take your own definitions of consensus and whatever, and see if your theoretical model fits as well the observations as mine. You take your Ptolemaic astronomy, and see if it fits the observations.
My description, as of this day, does: no economic parameter can be changed.
