It's impossible to really predict what that would look like but we'd have decades of research to work with before this became commercially available. Ideally you would engineer the organism to require a synthetic marker to reproduce that was tightly regulated and available only in the fertilizer so that it is incapable of spreading beyond runoff.

However, we're talking about a hyperthermophilic bacteria that wouldn't be able to out compete it's neighbors at temperatures anywhere near ambient. Realistically, the direct evolution experiment will likely require several genetic transfers of the chemical pathway to organisms better adapted to the environment so we won't be effecting survivability as much. Directing the evolution of a single feature is much more practical than the whole organism so I don't think we will be tipping the scales too much.