jackson-1 VS Scrutor

For some of the cases where type erasure would be a problem, there exists workarounds such as https://github.com/FasterXML/jackson-1/blob/master/src/java/org/codehaus/jackson/type/TypeReference.java anyway, so breaking the world to avoid erasure (Scala will be very unhappy if you remove erasure) seems like a bad trade.

You can with MS's default DI as well via scrutor.

if you are heavily invested in autofac you can replace the build in DI with another container. at this point i don't think it is that necessary. especially if you use scrutor. it gives your some nice assembly scanning registration that autofac/simple injector give https://github.com/khellang/Scrutor

If you wouldn't mind, I have one additional question. If I wouldn't bother to remove this loading, how could one add that as service when using DI? I have found tool called Scrutor (https://github.com/khellang/Scrutor), is it something commonly used in situation like this?

I much prefer using Scrutor and asbly scanning myself. https://github.com/khellang/Scrutor

you can also use Scrutor to get closer to Autofac functionality https://github.com/khellang/Scrutor

That can't be the only justification, because you can easily register services using reflection without using MediatR.

I might suggest additional registration time extensions to modify the behavior as needed on classes. Take a look at Scrutor for some inspiration on API specifically.

There is a package that simplifies it for you if you want. But not everybody agrees that automatic registration is good pattern, so you have a choice.