Dear Lawrence,
Interesting calculations. But my substitution (r, t) --- > (tau, xi tau^4) is suited for the Schwarzschild coordinates, so I don't expect them to work as well with the Eddington-Finkelstein's or to Kruskal-Szekeres coordinates, which indeed have the 1/r blow up.
As I said, in differential geometry it is known that you can't cover any surface with only one coordinate system. For example, for the sphere you need at least two coordinates, otherwise you have coordinate singularities. If it is impossible to cover the sphere with only one coordinate system, this doesn't mean the sphere is singular. It just means that it has different topology than R^2. In general relativity the things are more complicated, for two reasons: first, we deal with 4 dimensions. Second, the metric is indefinite, and it imposes constraints, in addition to those imposed by topology. But, if we can find, for a given spacetime, two or more coordinates which cover it and generate an atlas, and in which the singularity problems are resolved, the problem is solved. So, for r=2m there are the Eddington-Finkelstein, as well ass Kruskal-Szekeres, and Gullstrand-Painlevé coordinates, which preceded all. They work only for the event horizon, not for r=0. Mine just repair the metric at r=0, in the sense that they make it benign. They apply for r