Engineers have developed many different techniques to allow what is in effect lower ISO in extreme conditions. For example, one frequently examined technique is to have a pixel with a bilinear response. It would have normal sensitivity for most of its dynamic range, but as the amount of light begins to approach the pixel's "well capacity," the pixel would switch to a second, more gradual response curve (via a technical trick) that would prevent saturation in much stronger light.
Another trick is to use a logarithmic response curve, which accomplishes the same thing but without a sharp discontinuity (but with an over-all non-linear response).
People have also made chips with what are essentially neutral-density filters built in on sub-pixels, which can be used when the primary pixels are saturated.
Again, there are many tricks for doing this. However, for general photography, the "keep it simple, stupid" approach seems to be best. Therefore, you rarely see these tricks actually used. The major exception is the so-called "super ccd" approach, with a sub-pixel per pixel that retains responsivity in high light. This comes with a trade-off in resolution, however, and so it remains a niche product.
Cheers.