The only thing "raining down" will be a mostly intact RTG or intact GPHSes with no release of PuO2 in either case. They are designed to survive launch accidents due to the obvious concerns you express. The government doesn't let you just launch nuclear material without extensive engineering to deal with possible accidents.
From the Mars 2020 EIS, 4.1.4.3. MMRTG Response to Accident Environments:
- Explosion Overpressure and Fragments: Liquid propellant explosions and resulting fragments are expected to damage the MMRTG, but not result in any release of plutonium dioxide.
...
- Reentry: Impacts resulting from reentry of the MMRTG are dependent upon when and from where reentry occurs.
- Most suborbital reentries are predicted to result in intact impact of
the SV due to the presence of the SV aeroshell for Mars entry.
Releases in these cases are similar in nature to those from SV impact
near the launch pad.
- Reentry from circular orbital decay or long-term reentry is predicted to cause breakup of the SV and the MMRTG with subsequent release of the GPHS modules. (This breakup of the MMRTG and release of the GPHS modules is intentional and designed to limit the release of PuO2 in this type of accident.) This will result in some heating and ablation of the surface of the GPHS modules, but no containment failure or release in the air. When these separated components impact land, there is a potential for release from the GPHS module if the impact is on rock or a similar hard surface. No release is expected from a water impact or soil impact.
A very high-speed hyperbolic reentry could release PuO2 from a GPHS and disperse it in the atmosphere. That is not considered to be a desirable outcome, as implied by your question. There would be the possibility of excess cancer fatalities in the human population of Earth from inhalation of the resulting particles.
This had to be dealt with for Cassini, which had a 19 km/s Earth flyby on the way to Saturn. A great deal of work was done to assure that the probability of an inadvertent reentry due to any cause, including spacecraft failures, was less than $10^{-6}$. Furthermore, the project had to convince expert, non-advocate review boards (i.e. people who didn't care if your project went forward or not) that that had all been engineered and calculated correctly.