In mains applications with high voltage, high inrush currents, and nasty spikes, yes, driver failure is a significant failure mode. Inside a PC has less of that (though they’re still not immune to driver failure). Driver failures are usually near-total - a total failure or flashing/cycling are typical. These are very noticeable.

It’s not the only failure mode. Degradation over time is much less noticeable, because all lights located together typically have the same brightness. You only notice when one is replaced, or there’s a set of lights switched separately with a much lower hour count.

Many modern high-end LED drivers actually have programmable options to reduce the impact of this - for example, starting at 70% current and gradually increasing as the hour counter increases and the LEDs age to maintain constant brightness, or deliberately under-driving the LEDs when a replacement light is installed so as to match the older fittings surrounding it.

Operating temperature is one concern, but also simply current per die area. A 1W LED driven at 1W and heatsinked for 1W simply won’t maintain brightness as well as a 3W LED, also driven and heatsinked for 1W.

Different colours, chemistries, and manufacturing processes also have a big impact. A lot of R&D has gone into longevity and efficiency for white illumination LEDs - less so for each individual colour. Some colours naturally just don’t last as well - generally, shorter wavelengths (i.e. blue) degrade quicker. UV LEDs are often considered consumables due to their very short lifespans.

I’ve seen a number of routers and switches where the LEDs on ports that have been in use for years are noticeably dimmer than on mostly-unused ports.