I personally believe XT's (and as suggested here BU) 75% threshold to be dangerously low as well. A 95% supermajority with a minimum 2 week grace period and alerts sent should be the default for hardforks.
I think 60% is fine. Just kidding. Or am I? I say it doesn't matter what I think. The market will make the choice anyway. The only thing devs can do by locking the consensus parameters down in each implementation is force the market to make a less granular choice.
For example, if the market wants a 90% supermajority, and the options are Core at 98% and XT at 75%, maybe the market will be forced to go with 75%, which the market deems dangerously low but better than an impossible 98%, or it will be forced to go with 98%, causing a lot of unreasonable delay.
Or maybe it's Core with 95% and a three-day grace period, and XT with 75% and a three-week grace period. Gotta choose the lesser of two evils, not a happy place to be.
When choices are not granular, consensus cannot find its optimum level. This is the kind of friction that can be removed by unbundling the consensus-parameter-setting service from the rest of the roles the devs play. The choices can even be among BIPs the devs offer, so it's not even anti-dev. If the market thinks devs know best, the market will go with a dev recommendation. If not, it will go with something else.
And small block adherents, imagine the reverse, if it were Mike and Gavin were running Core and Pieter, Wlad, and Maxwell had broken off and started their own implementation, with maybe Jeff going between. And people were sticking with Core and its giant block plan, heading for catastrophe. You might notice the market friction then.
One doesn't have to pick sides. I respect all of the developers above and can have nuanced opinions and disagreements with individual aspects of their code contributions.
So do I. They all have great talents to contribute, and it pains me to see them have to be involved in a kind of power struggle because of the fact that they are being relied upon not just to recommend consensus parameters from among controversial choices, but to attempt to force them on users or at least spoonfeed them to users because they believe that they need to do so in order to generate consensus. I think some of them are unaware, or have not considered the fact that consensus can be emergent. I don't even think Gavin has fully internalized this, even though he did agree to the idea three years ago.