You can try increasing the iterations to try to help the solver not explode like that (there’s a physics action for that).
Another thing you could try is a distance joint from the first link to the last to help it from not getting too long. This is a general box2d idea, not sure if that shakes out in construct.
Another idea is since box2d resolves joints with changing the velocity you could try adding some damping on the chain link objects.
Final thought as to why the chain explodes is when the physics object gets updated. The loop is basically behaviors->events->draw->next frame. Any changes to the physics objects happens at the behaviors step. So maybe the lurch is from the links being in different orientations when the joints are added.
What I did once to correct that is to create the links, wait till next frame (after the behavior runs again), then add the joints. Or something like that.