You're right, I shouldn't have used an absolute language. There is also the issue of voter apathy so yes, a 51% attack might be worthy of a try for a wealthy enough adversary.
In terms of the split, if the community broadly agrees that there is an attack, then I don't see it having a long-term detriment. However, as you said, if it is more 'stealthy' then it would be harder to detect and correct.
All that being said, I think there is a theoretical safety mechanism in place. Whether that will work in real life or not, we'll wait and see. I am sure some good lessons will come out of this in either case and the next generation of applications can improve upon some shortcomings.
In terms of the split, if the community broadly agrees that there is an attack, then I don't see it having a long-term detriment. However, as you said, if it is more 'stealthy' then it would be harder to detect and correct.
All that being said, I think there is a theoretical safety mechanism in place. Whether that will work in real life or not, we'll wait and see. I am sure some good lessons will come out of this in either case and the next generation of applications can improve upon some shortcomings.