Intentional forking as a feature is ofc still possible. Validators owner who disagree with an amendment to the point that if it is accepted they would want to stay on the "old chain" and thus fork it, can simply remove the "yes" votes from their validator list essentially ignoring their votes and thus separate from each other. By the time one cluster enabled an amendment and the separated cluster does not, a fork happens*. This requires some kind of coordination and is unlikely to happen unexpected. A largely unpopular amendment would simply not reach the 80%. And one that reached 80% is unlikely to be consider so bad by the remaining 20% that they would want to fork. But the possibility is there it just never happened in the 8+ years its running.
*ofc it would also go the other way around. An amendment that wont reach 80% could also cause a cluster to intentionally separate and enable it there which would also case a fork.
The question was: "What exactly is the "threat" of forks?"
Your answer seems to be the answer to the question "How can we prevent forks?"
"Uncertainty" is certainly getting at some sort of answer, but falls short as you're not really explaining what kind of uncertainty and why it's bad in the first place.
