The statement "I've found that Aberth performs worse [than Weierstrass-Durand-Kerner] on most examples as its faster convergence is offset by the slower evaluation." is interesting.
Fredrik Johansson is the author of https://mpmath.org/, https://arblib.org/ and https://flintlib.org/ arbitrary-precision libraries and of the Mathematical Functions Grimoire https://fungrim.org/
https://flintlib.org/doc/examples_arb.html#pi-c has an example where π is calculated to 10^6 digits.