If you do much urban riding, you've probably encountered one of the practical cyclist's banes: embedded loop sensor activated traffic signals. Which is to say, loop sensors that ignore (or rather, fail to detect) bicycles.
There's one on my daily commute home. If I'm a little late on the way home and motor-vehicle traffic is diminished, I can wait through a couple of cycles. Nothing is more frustrating.
Apparently, this is a widespread problem for cyclists. There have been some gadgeteer approaches to this problem involving gluing very strong magnets or large metal plates to the bottom of your cycling shoes. Hm. I carry a laptop in my left pannier, and I'm reluctant to expose it to strong magnets.
So what to do? I've done some research, and have some information to share. I've come to the conclusion that this is a three-stage process; Keep advancing until you find a satisfactory solution for you.
1. Optimize your placement. First and foremost, know where to put your bike on the loop. (This is where I was going wrong.) The green zones in the diagram at right are optimal. (I was using a "Dipole" position on a "Quadrupole" loop. If you are seeing a "Diagonal Quadrupole" loop, you probably live in Davis CA, Boulder CO, or Portland OR.)
2. Know when you can "Cheat". If you use optimal position as noted above, yet you pass through an entire cycle and the traffic signal fails to let you through, then the signal can be considered "defective". Most jurisdictions allow running the light in such a situation. If you do this, be doubly careful! Watch most carefully for traffic that doesn't see you!
3. Contact your local Traffic Engineers. The good news about inductive traffic loops is, they're easy to adjust. The bad news is, you may get ignored by your local traffic engineers. When I emailed my county engineers about the light I'd been having trouble with, their advice was to dismount, walk as a pedestrian across 3 traffic lanes, and actuate a pedestrian call button. I hope you have better luck when and if you have to call them.
Here are some more avenues for research, if you want to dig deeply into the physics of embedded loops:
Goodridge article, "Detection of Bicycles by Quadrupole Loops at Demand-Actuated Traffic Signals"
Tracy-Williams article, "Traffic Signals"
John Allen article, "Traffic Signal Actuators: Am I paranoid?"