For example, to impede escape through pollen, a transgene could be inserted into chloroplast DNA, which in many crops is transmitted mostly through ovules, not pollen. Any of these strategies would be leaky to some extent.
In a paper appearing soon in Ecology Letters, Haygood, Ives, and Andow mathematically analyze how the probability of escape within any given period of time depends on several factors, including a leakage parameter characteristic of the strategy. They find that even when the average escape time is fairly long, there is substantial probability of escape much sooner.
Even with a leakage parameter as small as one in a thousand, the probability of escape within as few as ten generations could be appreciable.