Investigate handling of sub-command failures #84
Description
See #81 for an example that was leading to Aspire hanging. We need to confirm that we’re failing correctly on errors and that Aspire is similarly able to detect those failure states. This is complicated by the fact that we intentionally fork DCP out of the app host process tree to avoid getting shutdown by IDEs/debuggers for our graceful cleanup flow. It’s quite possible there are some points in the app host flow where we’re not properly surfacing DCP connection failures. We need to investigate both avenues.