Thank you for making me think about this.
In a real electromagnet, the ultimate strength of the magnetic field developed is a strong function of the amount of space filled by the wiring, the geometry of the armature, and the power dissipated in $I^2R$ losses in the wiring. If you take a motor whose field coil is packed full and rewind it with wire that has four times the area, then as long as you pack it full you'll use 1/4 the length of wire. The resistance will go down by a factor of 16 ($4^2$).
To get the same magnetic field, you'll need to drive four times the current -- which means that your "new" $I^2$ is 16 times as much as your "old" $I^2$. Since your new resistance is 1/16 as much, the overall $I^2R$ losses will be the same.
Which means -- on a basis of equal heat generated in the field winding, as long as you have a source of power for the field coils, it doesn't matter if it's series or shunt.
I think the reason that shunt windings are good for starter applications is because you get more heat in the windings than the motor can stand for long periods, but the amount of heat in the field winding is self-regulating -- it's only super-high in that (hopefully short) period of time when the motor isn't moving; once things get going, the field winding current just naturally goes down, without needing the intervention of humans or even automatic circuitry.