So it bursts into flame after ten years on the pole?

Sarcasm aside, it's one of those odd design corners where you really gotta respect the simplicity and dumb brute force robustness of a mechanical solution. It solves multiple problems at once: a time constant, accurate enough response, strong switching (hysteresis) with low resistance (mechanical contact), long life, and few quirks and other downsides (couple watts idle power, undesired behavior when cold or restarting). And failsafe (better to have lights always-on than always-off). And it uses as many mechanical parts, or maybe fewer, than an electronic solution, and being stamped metal and rivets, a fraction of the cost (well, maybe, or potentially).
I wonder what the transient immunity of CdS actually is; you wouldn't expect them to be very linear, at least at some point, and yeah I'm pretty sure you'll get avalanche at some point; but for a visible-width track, it's got to be pretty high... They're surprisingly linear, at least at what voltages and time scales I've tested them at, and seen applications for.
I don't think there's a true non-mechanical pure-semiconductor solution that strictly (all parameters) outperforms such a device, with less than... maybe $100 of parts? Depends on how much TVS is required, and Rds(on) say using SiC MOSFETs as SSR.
Maybe not so bad if some assumptions can be made about the load characteristics (surge immunity is a heck of a lot easier into an inductive load), and strict outperformance surely isn't needed in a real product, but it's interesting to illustrate just how well a mechanical solution can do.
Tim