It might or might not. This is a costing exercise where you need the understand the objective and cost factors.

If you have a strong requirement to avoid failure, proactive replacement let's you optimize the unit cost of replacement because you can fully utilize the crews and have no wastage.

If you dispatch reactively, you either have people sitting around waiting for light bulbs to burn out or lightbulb replacement pulls crews off of other work, impacting schedules.

...or you could hire contractors when you need to do bulb replacement? Then you have full utilization while you're replacing bulbs, no one is idling on your payroll while you're not, and no one gets pulled from their existing work to do replacement.

That's an option too, but is more complex and less deterministic.

The key assumption is you want low/no downtime. The problem you have is negotiating SLA penalties are difficult to get right for organizations like cities and getting good metrics are tough.

The other thing, especially with government contracting where RFPs are extremely difficult to change after the fact without getting contracts disputed by the losers is that the contractor will add risk premiums for penalties.

... and the external contractor who you're pushing all that risk on will charge you appropriately.

Sure, but they only receive the full amount of the risk you lose if you are their only client. They're still cheaper than either solution I responded to.

In the UK the device managing the junction can already detect most failed lamps, it's common in safety applications to have paired filaments, so when one fails the other still works but you can detect the problem. People selling a city devices to manage their traffic offer the feature of the city being informed about failed lamps. But, that costs money and saving money is important in most places, so they'd rather wait for motorists to report a failed junction and repair it after the fact versus paying extra to solve it first.

As things move to LEDs it hardly matters anyway, a correctly engineered LED lamp doesn't suddenly fail like filament lamps, it just gets sort of dim gradually, so a regular schedule (e.g. replace all the LED lamps every 5 years) is the correct way to deal with it. Cheap consumer-grade LED modules sometimes fail suddenly due to over-heating or badly made electricals but the road maintenance people probably aren't buying unlabelled stuff off eBay so they're unlikely to have that problem.

Replacing a traffic light after it already failed is too late.

Not in London. Nearly every day I see a new failed one. There is usually redundancy tho

That seems like the right approach. The traffic light has a backup bulb and alerts when one burns out. I mean, cars can tell when a bulb is out, right? Why shouldn't traffic lights.

Actually it's even simpler than that, just have a minimum of 2 lights per direction of travel.

So if one fails, a bulb burns out, one gets destroyed in a storm, or anything else happens, there will be at least one full backup.

Having automated outage reporting still makes sense though. It keeps you from having a guy that's paid to check every light in the city manually. And since your lights are on a city-wide network (they are, right?), then adding outage reporting should be a feasible thing.

Hey, that's a good idea - exception catching for traffic lights