I now have data files of 2,959,301 Ulam numbers <= 40,000,000, if anyone wants a copy them. One file is a text file with one number per line, 28MB (7MB zipped). The other has the number and the indices of the two Ulam numbers that make up the sum in a binary data file, 35MB (14MB). I can zip these if needed (recommended) or possibly get them into a different format (e.g. gaps, binary file of numbers only). Still no gaps larger than 587 and the density is holding up around 7.4%.