DNS, again you assume a person has been standing around long enough to read signs. Again, I say a person that observant, who is there at the start of this scenario, should have interceded as soon as the mother started placing the child on the rail. So, either your theoretical observer is not that observant, or else your theoretical observer did not arrive early enough for the luxury of time to read signs.
Or else your theoretical observer is the type to read signs while toddlers are under attack...
Now, I am in Afghanistan at the moment, and MWR bandwidth does not support video that well, so I did not see 11 dogs because I did not see the video, and the articles I read did not mention numbers.
With 11, I would shoot a few first, assuming one can carry at the zoo, and drop in after they started to scatter and retreat. Take them out of feeding mode, and into flight / retreat mode as it were. As others noted, gunshots plus yelps and fear pheromones would probably create more panic than gunshots alone. But, ultimately, somebody has to get in there and see if the kid is alive.
In my mind, based on both crowd studies and past experience with crowd reactions to dangerous events, if I do not take the lead - the odds are we will wait for official first responders. Ergo, I am in the habit of taking the lead.
I get that others look for reasons not to act, instead of means and methods by which to act. Really, I do.