The sensors work when there's a disturbance in the (magnetic) force caused by 2 tons of steel. Bike's are barely a quarter that, and not as spread out. I knew someone who was an avid cyclist, and did so in groups. They tried to pile their (assuming aluminum) bikes on the sensor with no luck.
Interesting solution with the magnet and starter.
Not all sensors are the same, but many of them are based on current detection. Under the road is a coil of wires with alternating current passing through. When nothing is on the road, the coil of wires is similar to the primary side of a transformer with nothing connected to the secondary. The circuit is self inductive and doesn't draw much current. When you put a large amount of inductive material on top of the sensor, it creates another path for the magnetic field to go; your car is like the secondary side of a transformer. Soaking up the magnetic field reduces the amount of self inductance on the primary side, which makes the circuit draw more current, and current is what trips the sensor.
So basically the question is how to create a magnetic short circuit. Instead of using a material that conducts electric fields (copper), use a material that inducts magnetic field (iron). Aluminum bicycles won't work because aluminum is a terrible inductor. Iron is great, nickel is good, cobalt is ok. The construction makes a difference as well. High quality inductors found in power transformers are made of multiple layers of iron that are magnetically separate from each other because the stranded design works better than a solid design. Soft iron is a better inductor than hard iron.
Permanent magnets on the bike might work depending on whether the sensor is looking for RMS current or peak current. Permanent magnets have one direction, but the sensor's field is alternating. Mixing a permanent magnet with the alternating magnet would mean the circuits impedance would go way up then way down, large spike up, then no current, then large spike, then no current. The RMS current would be slightly higher due to the large fluctuations, but the peak current would be much much higher.