The hydrogen bubbles are normal byproduct of the chlorine generation and indicate it is producing -- in fact that's how I confirm at a glance if my SWCG is running. Different flow rates, lighting angles, eyeball position, pump flow rate, etc, can make them significantly more or less noticeable. If the bubbles stop/start when the only difference is the SWCG off/on, that's almost certainly what they are (a leak would be the the same on/off). A fun experiment is to take a lighter (the wind-proof torch style works best) when it's dark and watch the bubbles pop/explode.
I'd echo the suggestion to do an Overnight Chlorine Loss Test (SWCG off) to verify that you don't have something eating FC, and then after that you can do an Overnight Chlorine Gain Test (with the SWCG on) to verify how much it produces at night. Bring us the before and after FC numbers in both tests and we can help compute/evaluate.
What's your CYA level? That makes a big difference in how much FC is consumed by the sun, and it's quite common to lose CYA over the winter. It could very well be the difference between "FC remains the same or increases with X hours of runtime" and "FC goes down even with X hours of runtime"
