Incandescent lamps only represent a "sizeable" load on startup. Once the filament gets warm, current approaches that of a purely resistive load. And, as startup can (theoretically) occur at any point in the (half) waveform, the chances of the cold filament encountering PEAK line voltage is small. E.g., if the instantaneous line voltage is
20V, then even a 10X reduction in cold resistance makes the lamp look like a *warm* lamp seeing 200V (instead of 170V that it would normally see at peak).CFL/LED ballasts, OTOH, are *always* reactive -- rarely have power factor correction. So, that "surge" happens every (half) cycle. And, ALWAYS at the peak of the cycle (when the bridge is conducting!).
*Opening* the relay at this time usually poses the biggest risk for contact weld as current is already flowing and wants to continue to flow across the ever widening gap between the contacts. So, you draw an arc and start melting metal.Engineering is the endless pursuit of the "least bad" solution -- acknowledging that all solutions are "bad" in some way...