[j-conn]

Love this idea. My first thought is that calibrating mics could be difficult. I wonder how good existing decibel meter apps are

[Animats]

You locate using time, not amplitude, so microphone calibration isn't an issue. Take the incoming waveforms with close timestamps and convolve them on the server to line them up. Get ΔT values for each pair of phones picking up the matching sound. Each ΔT value places the source on a hyperbola. You need at least 3 phones to get a location where the hyperbolas intersect. More will be better.

Multipath always makes a delay longer, not shorter, which helps in resolving ambiguity in areas with reflective building walls.