Quite hard Form109, nothing is eternal! Even the LifeGuards from the 60s.
A bit corrected: They should not fail within well predictable (long enough) lifetime, so their mortality curve would be nearly rectangular. Then it's not as big issue to plan-in their group-replacement (together with relamping), so no photocell then fail on the street.
Loose welds is a quality problem (voids in the material, thermal stress) and this should be possible improve.
Eye wear is well predictable process, so it's lifetime should be well predictable and the eye designed so it will reliably work for the required lifetime.
Material fatigue is again well predictable process, so it's lifetime should be well predictable (contacts, springs).