/backport to rel/3.9

Started backporting to rel/3.9: https://github.com/microsoft/testfx/actions/runs/15442261447

/backport to rel/3.9

Started backporting to rel/3.9: https://github.com/microsoft/testfx/actions/runs/15443366680

Codecov Report

Attention: Patch coverage is 95.23810% with 1 line in your changes missing coverage. Please review.

Project coverage is 76.34%. Comparing base (35ac25e) to head (83f4e0b).
Report is 12 commits behind head on main.

@@            Coverage Diff             @@
##             main    #5694      +/-   ##
==========================================
- Coverage   76.39%   76.34%   -0.05%     
==========================================
  Files         600      600              
  Lines       36743    36767      +24     
==========================================
+ Hits        28068    28069       +1     
- Misses       8675     8698      +23     

Do you need a test for the situation that this is trying to "fix"?

If the binding redirects are messed up, it typically leads to end-less stream of problems. This workaround is likely just kicking the can down the road.

Do you need a test for the situation that this is trying to "fix"?

Probably, I can try to create a scenario where this is broken.

If the binding redirects are messed up, it typically leads to end-less stream of problems. This workaround is likely just kicking the can down the road.

I definitely agree. Once the user for example tries to actually use ValueTask as a return type of a test method (which is extremely uncommon), they will hit the failure to load the assembly again. However, it's not yet clear what's causing this in the first place, and it's more likely to be a user error causing binding redirects to not be correctly generated. I'm only trying to reduce the noise when users have an error on their side that causes binding redirects to not be correctly generated.

I'll get to writing a test for this later as few important stuff are showing up.