[Rafael-SOWNet]

Seven problems of the Bronstein, Hearn, Hebisch, Timofeev and Welz suites were declined or ran out of time. Six small changes answer them:

• A logarithmic part past the twelfth degree, up to the sixteenth, where every residue lies in a field of degree at most two. Hearn's degree-13 over degree-14 problem is over P^2 - 2 Q^2, and its residues are the roots of 4 t^2 - 4 t - 1, giving two logarithms of P ± sqrt(2) Q. A mod-p screen leaves the path before factoring wherever a factor of the resultant of degree three or more shows modulo a prime.
• e^(x + e^x) kept whole as a substitution's candidate. Hebisch's (1 + e^x) e^(x + e^x)/(x + e^x) is Ei(x + e^x).
• The root of a real square beside a constant off the real line, over a power of a real linear. The quadratic keeps the sign of its imaginary part on the real line, so the root and the artanh's logarithm stay continuous there (Welz 44).
• A rational function times sqrt(c x^2 + d sqrt(P))/sqrt(P), with P = a + b x^4 and c^2 = b d^2. It is written as two roots of quadratics (Welz 47 and 48).
• A symbolic power inside a reciprocal of a root. Timofeev's 1/((1 + x^(2n)) sqrt((1 + x^(2n))^(1/n) - x^2)) is arctan(x/sqrt((1 + x^(2n))^(1/n) - x^2)).
• A rational function of x and L = ln(x) over distinct linears in L, split into partial fractions in L. Bronstein's (2 L^2 - L - x^2)/(L^3 - x^2 L) becomes the logarithmic integral and two logarithms. This applies only for two linears or more, and only where nothing answers the integrand whole; it is dispatched at the very end.

Measured on 8a1ef454, branch against master:

Run	Problems	Master	Branch
Pocket: the independent suites and Rubi's 3.x	1,207	1,142 solved	1,143 solved
Family 1, 40 a file	1,381	1,362	1,362
Family 0, whole	1,814	1,789	1,796
Families 2 to 8, sampled	2,410	2,362	2,361

Re-run alone, nine moves are answered against three on master. Bronstein 22, Hearn 365, Hebisch 15, Timofeev 448, Welz 44, 47 and 48, and Rubi's 3.5:364 go from unevaluated to answered. Family 0 goes from 1,789 to 1,796.

Two problems master answers in 21–34 s, 3.2.1:292 and 3.5:37, timed out on the branch when re-run alone. At a 60 s budget both builds answer them, with the same allocation to 0.001% (34.2 and 37.0 GB), so the branch does not reach them. The one-row difference in the sample is 6.4.2:50, which both builds answer alone, in 18 and 21 s. No run had a wrong answer.

The rebase met two rules master had appended at the end of the dispatch; this one goes after them. SeamsOfTheIndependentSuitesIntegralTest differentiates the answers back, and the 533 related tests pass. BREAKING-CHANGES.md has the entry, with the 2.5.0 column measured on v2.5.0.

Part of #233.

🤖 Generated with Claude Code

https://claude.ai/code/session_012sonx8iAspMiwRwokT1Ura