Ahhh, but not quite...
I should have also been more specific with conditions (and reading back, my terminology was a little off)... Hopefully this makes more sense...
1. Supply voltage coming within about 2V of the set output voltage does indeed cause the output to deregulate. No mystery there.
2. Under current limit conditions, lets not forget that the current sense happens on the supply side, and that this is of course a buck converter. So if a load is present that is causing the 300mV threshold to appear across Rs, a change in input voltage will have no effect on this drop provided the load continues to pull sufficient power to keep it there.
What this translates to is, an increase in supply voltage would raise the input power, which gets passed along to the load because the voltage at the input is being converted to current at the output. Triggering the 300mV at the input would have already caused the output voltage to dip, but now there's more current flowing in the output, which would tend to push the output voltage back up toward the set level. So, the load will then be receiving this additional power that is available at the input.
3. The chip however remains protected, because the current that causes the voltage drop across Rs is equal to the current in the switching transistor, as both are before the inductor that does the actual conversion.
4. The output current would rise with increased input voltage until the output voltage was sufficient to reach the 1.25V threshold.
5. I agree, there's a bit of a paradox at this point, and without actually building a test rig to see exactly what would take place next, I can only take critical thinking so far to determine the outcome. My thinking is, ohms law must still be in effect, and so the voltage at the output must rise until the chip backs it off as part of its limiting system, but since limiting is already taking place (via 300mV across Rs), it isn't going to want to allow the output voltage threshold to be hit to do its part of the regulation. Without a further increase at the output from either angle, both would stabilize, however this is illogical, as there is an increasing amount of power available at the input, and the load isn't going to magically "back down" (unless it pops).
You've got me really curious at this point what exactly would take place under these conditions, but while I have the equipment to do some testing, I don't have another converter handy that I can experiment with at this moment. I likely have a few more somewhere in storage, and when I get them out, I'll be more than happy to play with one and plot out the behavior to answer these questions once and for all.
Of course, if anyone beats me to it, that's fine too. The conditions are pretty much described above. Using a higher value Rs and a simple resistor for a load should reveal whatever there is to be seen, all within the safe limits of lower power levels.
Steve