Partial answer addressing Q3 for SpaceX Dragon2 missions. Extensive leverage of my answer to Dragon re-entry flight profile? & my answer to What is the heat shield refurbishment procedure for a crew Dragon capsule?.
There are 7 normal splashdown sites all located near northern Florida for crewed flights:
They are not appreciably far enough apart to warrant distinction when considering the relatively low accuracy of my "armchair" modeling.
Given the 51.6° inclination of the ISS there are 2 possible tracks that a returning Crew Dragon could take, north(L) and south(R):
The track of choice for a given splashdown time can be inferred by using ISS TLEs to generate a "full orbit" ground track plot at the splashdown epoch (no SGP4 needed, mean elements suffice!). Then simply see which way the ground track crosses northern Florida:
(Personal work, CRS-23 example)
I know the Demo-2 mission used the southern track, and some crude TLE work (like above) shows both Crew-1 & Crew-2 used the southern track, while CRS-23 (the October 2nd example in the question) used the northern track as shown above. Note how even 5 days out it is easy to see which track is going to be used.
I think it's safe to assume that the higher the instantaneous reentry heating, the more luminous the capsule will appear in the sky (at the very least it's a good approximation). Here is the entry trajectory where the heating rate is expressed as a colour using MATLAB's "hot" colormap (lighter = hotter, linear scale):
(Personal work, heating estimated from here)
This looks promising considering the news reports from Alabama & Georgia given in the question and the peak brightness occurring over those states. Here is a plot of heating VS altitude, the higher the capsule is the farther away it should be visible from:
The time from entry interface until parachute deploy is about 12 minutes and estimated splashdown times are usually announced.
SpaceX also does west coast splashdowns of some CRS missions near Los Angeles (iirc):
People fortunately located near these tracks should be able to reliably predict and see SpaceX capsule reentries.
I have created KML files (Google Earth Pro, it's free!) for the 4 trajectories, found on Github. This is powerful for observation planning. For example, here is the view from the US Space and Rocket Center in Huntsville, AL for a Florida north track trajectory (moves right to left):
This is also near peak heating so it would make for a very cool photograph/video (sadly no longer in HSV or else I would try this!)