InvalidProgramException in 1.9.0 on .NET Core 2.2

Describe the bug API throws an exception rather than completing request

To Reproduce Steps to reproduce the behavior:

1. Do lots of API requests, and eventually there will be a period of time where many of them throw this exception

Expected behavior Exception is not thrown, and API requests completes normally.

Runtime environment (please complete the following information):

• Instrumentation mode: manual with NuGet package
• Tracer version: 1.9.0
• OS: docker image microsoft/dotnet:2.2-aspnetcore-runtime in EKS kubernetes cluster
• CLR: .Net Core 2.2

Additional context Stack trace:

System.InvalidProgramException: Common Language Runtime detected an invalid program.
   at Microsoft.AspNetCore.Mvc.Internal.ResourceInvoker.ResultNext[TFilter,TFilterAsync](State& next, Scope& scope, Object& state, Boolean& isCompleted)
   at Microsoft.AspNetCore.Mvc.Internal.ResourceInvoker.InvokeResultFilters()
   at Microsoft.AspNetCore.Mvc.Internal.ResourceInvoker.InvokeNextResourceFilter()
   at Microsoft.AspNetCore.Mvc.Internal.ResourceInvoker.Rethrow(ResourceExecutedContext context)
   at Rethrow(Object )
   at Datadog.Trace.ClrProfiler.Integrations.AspNetCoreMvc2Integration.Rethrow(Object context, Int32 opCode, Int32 mdToken, Int64 moduleVersionPtr)
   at Microsoft.AspNetCore.Mvc.Internal.ResourceInvoker.Next(State& next, Scope& scope, Object& state, Boolean& isCompleted)
   at Microsoft.AspNetCore.Mvc.Internal.ResourceInvoker.InvokeFilterPipelineAsync()
   at Microsoft.AspNetCore.Mvc.Internal.ResourceInvoker.InvokeAsync()
   at Microsoft.AspNetCore.Builder.RouterMiddleware.Invoke(HttpContext httpContext)
   at Elmah.Io.AspNetCore.ElmahIoMiddleware.Invoke(HttpContext context)

Packages related to Elmah:

  <ItemGroup>
    <PackageReference Include="Elmah.Io.AspNetCore" Version="3.7.82" />
    <PackageReference Include="Elmah.Io.Client" Version="3.6.73" />
  </ItemGroup>

Mentioned this in #541 after I upgraded to 1.9.0 hoping this issue was fixed. We currently have DataDog tracing disabled in our production API until the issue no longer happens in QA. Maybe related to #564

[fjmorel]

I haven’t had this issue since forcing the docker image to use 2.2.8 instead of whatever 2.2 version it had cached. I tried that after seeing comments in #564.

[andrewsouthard1]

Hey all - we expect this issue to be resolved in later versions of the tracer. We’re going to close this issue but please do reach out again if you run into any problems!