Bump! I have been wondering something on the same lines. I don't have multiple audiomoths. 

Before chaining them together physically, I would try using the Audiomoth Dev connected with the Audiomoth GPS Board (note there are options for just GPS and for GPS + GLONASS). If you use the Audiomoth GPS Firmware, they will automatically sync time to a reasonable degree - I think < 1 microsecond. I have used the Devs in a wide array (units spaced by ~50m) with decent accuracy for sources within the bounds of the array. The time error was relatively insignificant at that spacing, although if you are doing an array with the units very close together, the time error may become significant. I've been wanting to do a test with the Devs to quantify how close together they can be to accurately localize or get a bearing estimation using the GPS firmware.

For this to work, you'd have to know the distances between the microphones with extremely high accuracy if they were close. I don't know if mounting to a peice of wood and measuring the distances would suffice. I would have a machined piece of equipment with mounting posts for each Audiomoth that were at known distances with high precision. Maybe a laser range finder could work though.

Another more off-the-shelf option would be Frontier Labs BAR-LT units, which come with software for localization. This link also explains how to do an acoustic calibration with a sound source at a known location (similar could be done with Audiomoth). The Audiomoth case + Audiomoth Dev + GPS board is ~$210 USD while the smaller Bar-V2 runs about $320 USD. The developer Michael Maggs is very responsive and can give more information.

One issue I would anticipate, is that Audiomoth does not provide complete Access to I2S, or do you mean I2C, which is much slower, obviously.