It is also possible to re-tune the bacterial ecosystem in our favor. First, we can introduce predators. Vaccines, when available, can prove to be very succesful. The Hib (Haemophilus influenzae) vaccine reduced the number of cases by two orders of magnitude, and along with it, bacterial meningitis and pneumonia fell as well.
Phage therapy, the introduction of antibacterial viruses, has entered clinical trials with initial success. Phages can even be used in sutures and dressings and can be engineered to light up (using luciferase genes) when they have infected bacteria!
The other option is to modify the competition. Instead of trying to kill the offending bacteria, treat the human body as a natural habitat for bacteria, most of which are harmless. Since virulence factors are not as important to survival as, say, ribosomes, the selective pressure favoring any resistant bacteria will be much smaller.
One interesting target for inhibition is quorum sensing (QS), the bacterial communication necessary for coordinated attacks (as in cholera) or biofilm creation (like in chronic Pseudomonas infections). Several strong QS inhibitors have been found and demonstrated to work against biofilms and toxins, though, trials of one, azithromycin, were canceled due to funding problems.
Finally, it might also be possible to introduce harmless bacteria to crowd out the pathogenic ones, as was done in trials of an engineered nontooth-decaying S. mutans strain.
Bacteria will always find a way to surprise. In the azithromycin QS inhibitor trial, while the wild-type expressed less virulence, a population of freeloading QS-oblivious, less-virulent mutants also decreased. So while you might see some success with the “harmless competitor” approach, it could be negated by a QS treatment.
Unpredictable as the process might be, we must continue to develop new approaches to controlling microbial infections, at least to account for that which policy cannot. That boulder isn’t going to get any lighter.