Polled "congress approval" has no applicability, since we have direct popular election of both senators and House representatives (and, here, of state senators and representatives in Florida). The Constitution includes no direct provisions allowing disapproving citizens in California to disempower representatives in Maine (or, presumably, though I've never read the Florida Constitution, of residents of Tallahassee to disempower representatives of Miami).

Pretty sure the evidence is quite clear.

The first black president was elected promising change.

Now we have Trump who had no political background implying people want change

Choosing the least painful way to get screwed isn't consent.

assume the US voting populace was two sets of clones, 50% Mitch McConnell Republicans and 50% Nancy Pelosi Democrats - but the distribution always varied within a standard deviation (eg year to year or district to district or state to state). Maybe there's even some generational biases, where one set of clones is more common for a decade.

I think you'd find that pretty often, more than half the country disapproved of congress, even though, in the long run, each clone is equally likely as a voter regardless of geography, economic interest, etc. How could you design a voting system that's responsive to these variations?

This challenge gets more complex the more political alignments you account for.

Few people in an opinion poll of Congress are assessing their love of bicameral legislature. More commonly they're stating that they wish they could just have their way, regardless of opposing voices.

Special pleading aside, elections are how we gauge consent. What would be coercive would be if we deputized firms like Gallup and made their polls binding.

But putting principles to the side: your argument doesn't even hold up empirically. Americans have a very low opinion of Congress, but generally approve of their own representatives.