In this paper, within the specific framework of a recently developed information flows preserving calculus of variations, we investigate a class of average fixed optimization problems, over sets of laws of Itô semimartingales. We show that, within those conditions, the optimums are clearly ruled by a least action principle, which yields the corresponding Euler-Lagrange conditions; this enlightens their specific dynamics. In particular, this encompasses a specific class of semi-martingale optimal transportation problems, and specific entropy minimization problems in close connection to the Schrödinger problem.