##F=ma## and ##F=dp/dt## are equivalent. Whether you mean acceleration or change in momentum, a force must act on the centre of mass of the system to change its state of motion.
There are many facets to this confusion but I think it mostly stems from the mistaken assumption that the conservation of momentum holds for magically 'disappearing' mass. It doesn't, because there is no such thing as simply disappearing mass. Mass (=energy) always leaves via some process, and carries away any momentum that the reminder of the system losses.
When you shoot a gun, the momentum of the bullet changes because there are forces acting on it, and similarly the momentum of the gun changes because there is an opposite force acting on the gun. Taken together the momentum is conserved, but the changes in the momenta of the components of the system are the result of concrete forces acting on each one of them.
If the net momentum of the lost mass is 0, as is the case with radiation (it has momentum too ) leaving the star isotropically, then the reminder of the system doesn't gain or lose any momentum. All the forces resulting from the isotropically leaving radiation cancel each other out.
Furthermore, as seen from the outside of the system, i.e., taken together the star plus the radiation it has so far emitted, its total mass remains unchanged (where we calculate the mass-energy contribution of the radiation).But if that is not convincing enough, maybe this will disabuse you of the idea: consider the mass lost in the lifetime of a star due to radiation amounts to less than 0.7% of its total mass (the mass defect in hydrogen fusion). The expansion of the universe is observed to carry both stars and inert (non-fusing) gas in the galaxies at recession velocities many times the speed of light.
This tells you that even if what you envisaged was a real process it could not account for the velocities observed,
and that it lies beyond the domain of applicability of simple Newtonian mechanics due to the velocities involved.
@Drakkith how is a star losing momentum (in some arbitrary reference frame)? See DelcrossA's post #12.