I've made some experiments on werecreratures, including with saves-scumming, and the transformations were always on the same tick. It isn't on midnight, it's on the same moment the moon changes to full moon. There are 13 transformations every year, more or less evenly spaced, with the actual calculated tick rounded up to the nearest tens - so different lunar months have different length, thought each specific one is always the same length, and begins in the same instant as in other years.
The transformed creatures have their target selected instantly, on the same tick as the transformation, be it another creature or a building, but the target can change on the next tick (for example from a building to higher priority live target, or from another werebeast to the baby carried by weremother). The initial target depends on the positioning, and it can be selected only in a time window. The target can also change during subsequent ticks, but after some time window they won't attack the same were-species.
If there are three werebeasts and a non-werebeast, and by chance the werebeast A and B attack and switch targets only between werebeast C and the non-werebeast in the beginning, then after dispatching of them, the werebeast A and B don't attack each other. If the switches are between them all in the beginning, then only one will remain (which can die of blood loss, eventually). If you teleport werebeasts from different cells to one, after this time window, they won't attack each other.
During tests I never had a situation that they won't attack someone, even if all were werebeasts. I've tested only werebeasts of the same curse. However, I use only relatively small cells (like 5x5), so everyone always sees everyone else in such cell, also for tests I was choosing the same transformation date (which may something to do with the length of this vulnerability time window).
Now if you have them in a small cell, and they won't attack each other months after months, but only start later, than would be interesting.