I can see why road maintenance hassle doesn't outweigh potential benefits, but what about a gene for producing a pesticide. Resistance to herbicide doesn't present an obvious fitness benefit to a wild hybrid, but not being eaten by bugs certainly does. How would said pesticide affect bee populations if all the wild relatives of a given GMO crop now produced its own pesticide?

Plants generally already do this. There are tradeoffs between productions of natural pesticides, rate of growth, tolerance to environmental extremes, and so on in plants, and our crops generally produce less pest-combating compounds than natural strains, in favor of greater growth rates, and we use artificial pesticides to make up for their weak natural resistance.

Our genetic modification technology is still quite a ways short of being able to design genes for the production of pesticides more potent than plants already produce, which will not divert significant resources from growth.