Investigate the curious case of GHC > 8.6 linker speedups

I found some magical speedup when gold is invoked on .a file GHC produced, with 20x less IO being done by the linker:

We should investigate whether that's an improvement in GHC > 8.6, or something due to the fact that I'm not using a bindist.

Maybe that speedup can be achieved in general?

To upload designs, you'll need to enable LFS and have an admin enable hashed storage. More information