If the constellation were a phased array, mutually coherent (easily implemented) each could beam at the same point on the earth, the interference pattern, i.e., the beamwidth of all of the elements would be proportional to the wavelength divided by the effective width of the constellation. Extremely small.
If one satellite failed the effect outside of the intended target area would be comparable what would be experienced if only that one satellite were radiating. A few million watts over an area the size New England would be a negligible health risk, microwatts per square meter. Each satellite could weakly modulate its signal to provide a health indication of the array with out substantially defocussing the beam. Current phased array radars with thousands of elements radiate periodically into a near field horn, modulating one phase shifter at a time to provide real time monitoring of the array status. It's like weighting a mosquito on an elephant, doable if you have a high enough SNR and weight your elephant often enough. The modulation waveform could be like a cell phone's CDMA waveform, a set of orthogonal signals, which can individually be picked out of the cacophony of competing signals.
If I read your post correctly, all the satellites would beam their energy at the same point on the Earth. I think your example of New England is entirely appropriate; I too would beam the energy there. I am not too worried about health effects, but I am skeptical about the practicality of such a power scheme. In your opinion, how large would the power collection antenna have to be? Would it be an array of antennas spread over some area of land? How efficient could this whole solar-to-radio-to-electricity thing really be?