I doubt it helps for CORE-5007 (as it was stated in master),
that ticket most likely got fixed beforehand already.
But I already tested: Delay-loading fixes CORE-15795,
which was unhidden by 0.4.12-dev-632-g
d93e516747
The fix by delay loading was cherry picked from commit
0.4.14-dev-475-g
82b758974f
It's not just a workaround, Windows delay-loads like that as well.
Thomas believes this specific delay-load will not trigger
the LD bug CORE-6504 within our gcc toolchain.
Thanks for help of Mark Jansen, Thomas Faber and Fabian Maurer.