Trigger runs for multiple stacks as a single batch with a shared dependency tree. For tracked runs, downstream dependents are resolved across all input stacks together, so a shared dependent only gets a single run. The output contains the runs created for the requested stacks as well as any runs created for their dependents, in no particular order — correlate results to the input via each run's stack rather than by position. A run's full details are only exposed when you have read access to its stack; otherwise only its id is returned.