Lucas Trzesniewski

> I'm assuming that the StructureMap build plans were cached for performance reasons and if so this will lead to a performance regression as the implementation in this PR is...

Also, what's the migration plan? I don't really like having to reference two distinct IoC containers. It would be really easy to mess something up down the road (libraries referenced...

Maybe we could reference only Lamar from Zebus, and provide a shim from StructureMap to Lamar internally (reimplement the part of the StructureMap API surface we use on top of...

> do you happen to remember which benchmark you ran to measure the throughput? It might be helpful to use the same one if you still have it. Sorry, that...

As this will be added to all messages, something short like a GUID should be preferred IMO.

Also, currently there is no way to configure multiple filters in `MessageDispatcher`, and if I want to force a specific `IMessageHandlerInvoker` to be loaded, it still needs to satisfy the...

This is absolutely on-topic, and could actually solve the issue I described above 😉

@ocoanet I started implementing something, so let's talk philosophy ```C# var fooTask = bus.SubscribeAsync(foo); // fooTask should not be completed before the bus starts, right? 🤔 // Should we make...

> It could be acceptable to return a completed task here because the subscription is already registered (locally) and it cannot fail That's true, and there are some advantages as...

I can see you're using a transient peer (the `*` in `.WithPeerId("BusMg.*")`). Commands are designed to be handled by persistent services, not by transient peers like the one you have...