This is not a very useful Twitter post at all.
The colours will relate to number of cases or number of cases per capita, but these are arbitrary values. Neither figure is necessarily more or less misleading than the other without contextual information for what the categorisation of the colours are and why those ranges were used. A range might want to give a good span of the data: the green map lacks granularity at the lower end, but the red one lacks granularity at the high end.
Let's imagine a hypothetical disease where there was a peak with up to 5000 cases per 100,000 people a week. Using a 3 colour chart, they set 0-1000 as green, 1001-3000 as yellow, and 3001+ as red. Then three months later, using this categorisation, it's well past peak and infections have plummetted so green abounds. So someone decides to change the ranges: this better represents the current infections, but it is arguably misleading people how severe the infection risk is, given that everywhere it is much lower than the peak.
So what is the right thing to do? That's not an easy answer. Trite Twitter posts like that don't help enlighten.
One can argue that colour selection also matters for the visual impact (red = alarming, green = soothing). But if that is so, a graph that goes blue-yellow-orange-red is weighted towards making someone feel worried; why is this better than green-yellow-orange that's weighted towards making someone feeling reassured? And again, see the issue with granularity above, which accentuates the issue.