Compare the cost of stopping to the cost of a launch that should not have happened. A stop costs the time already invested and whatever it took to reach the conclusion. A failed launch costs all of that, plus the inventory, plus the packaging run, plus the retailer relationship that took the hit when the product did not sell through.
The asymmetry is large and well understood, and teams still find stopping hard. Partly that is sunk cost. Partly it is that nobody gets promoted for the launch that did not happen, so the incentive points one way while the arithmetic points the other.
What helps is deciding the stop condition before the evidence arrives. A team that has written down what result would kill the project is far more likely to actually kill it than a team evaluating the same result with no prior commitment.