SkipOneActivityPropagator.cs

DistributedContextPropagator.Current = new SkipHttpClientActivityPropagator();

public sealed class SkipHttpClientActivityPropagator : DistributedContextPropagator
{
    private readonly DistributedContextPropagator _originalPropagator = Current;

    public override IReadOnlyCollection<string> Fields => _originalPropagator.Fields;

    public override void Inject(Activity? activity, object? carrier, PropagatorSetterCallback? setter)
    {
        if (activity?.OperationName == "System.Net.Http.HttpRequestOut")
        {
            activity = activity.Parent;
        }

        _originalPropagator.Inject(activity, carrier, setter);
    }

    public override void ExtractTraceIdAndState(object? carrier, PropagatorGetterCallback? getter, out string? traceId, out string? traceState) =>
        _originalPropagator.ExtractTraceIdAndState(carrier, getter, out traceId, out traceState);

    public override IEnumerable<KeyValuePair<string, string?>>? ExtractBaggage(object? carrier, PropagatorGetterCallback? getter) =>
        _originalPropagator.ExtractBaggage(carrier, getter);
}

I was happy with this workaround so I could create a tree that was completely traceable without having to log all activities created by e.g. HttpClient.

However, using this in .NET Isolated functions with Application Insights will make the application map incomplete. I had a case where I was making calls to CosmosDB which only became visible after removing this workaround.
This workaround has its benefits, but the user should consider the trade-offs before implementing it.

@wkoeter would you be able to provide more details ?
What do you mean that this workaround caused application map incomplete ?

I would appreciate your expertise.