Something to keep in mind here is that we are looking at countywide data, which only goes from a bit under 1/2 the national median income in the poorest counties to a bit over 2X the national median income in the richest counties. A lot of the errors in contemporary political analysis come from treating 1/2 median income households as "the poor" and 2X median income households as "the rich" when the federal poverty line runs from about 1/5th to 1/3rd of the median income (approx. $12-20K) for most households and what most people think of as the cut off for "the rich" hovers around 5X median income (approx. $250K). So when we look at the county level, we aren't really getting a good sense of "the rich" or "the poor."
My sense is that "the poor" in the sense of the federal definition of poverty still vote 2:1 or 3:1 Dem consistently, while "the rich" in the sense of $250K+ never voted Dem since LBJ and then swung to a a near tie for Obama 2008, swung back to Romney 2012 and then swung really hard to Clinton in 2016. You have to look at the town or zip code level to see this, though. Think Atherton and Palos Verdes, CA. When you look at aggregate data, all you can see for "the rich" is the tail end of the $50-$150K GOP base, which greatly obscures the shift toward a $40-$100K GOP base under Trump.
That's true, counties are huge. There was a thread on how the wealthiest towns voted in 2012 and 2016. Places like Atherton and Darien are located in solid D counties, but they still voted for Romney in 2012 and they indeed swung hard to Clinton in 2016. 18 of the 27 wealthiest cities in Southern California voted for Romney in 2012 (I guess a similar number voted
Clinton in 2016). But the interesting thing is that Trump seems to have performed similarly to McCain with the very wealthiest voters.