When a motor is already rotating, the rotor's mass and velocity carries a momentum and will continue to rotate in one or the other direction to align to the stator's magnetic field as the polarity of coil switches periodically.
Let's consider the simplest brush motor with only one pole as shown in the picture. If the rotor initially sits at rest in perpendicular to stator's magnetic field before current is applied, then once current is applied the rotor can either rotate clockwise or counter-clockwise to align to stator's magnetic field. The initial direction should be indeterministic (or 50% equivalent chance of rotating either clockwise or counter-clockwise).
How are those small DC motors always rotate in one direction unless we reverse supply's polarity?
Most results I found on Google and videos on Youtube are self-contradictory or straight out erroneous.
Here I provide the diagram for Fleming's left hand rule for motor and the cross section of a 3-pole brush DC motor, where the green arrow indicates the close path of the magnetic field, and right arrow indicates direction of a conventional current. The coil wrapped around the pole (circled in red) is subject to higher current than other two poles (hence the bias). The magnetic field generated by the coil (circled in red) opposes that from the stator (dark vs. light green arrow). Rotor has to rotate then - but to which direction?
If I apply the Fleming's rule, the force induced by the front section of the coil will get canceled out by that part of the coil that goes behind that pole. Coil on the right will have a force pushing to the left (CCW), canceling out the force pointing right (CW) from the left coil. It's easier to visualize it through the diagram along with description given for F, B, R and L - all induced Lorentz forces get cancelled in that coil that wraps around the directly energized pole!!
The only explanation I can see for why the rotor has to "move away" from its current position is because the magnetic field generated by the coil (lighter green) goes in opposite direction to that of the stator's magnet. But to which direction the rotor rotates away should be indeterministic. In other words, if I tab the rotor CCW with my finger right before current is applied, the motor will rotate CCW (because I give it that initial momentum), or vice versa for CW rotation.