Whether jobs are started asynchronously by sp_start_job or sqlagent's call to it, there is a per-sqlagent-subsystem max worker thread limit. Once that limit is reached, SQLAgent.out logs that stall. Best to inspect sqlagent.out while racking up a number of concurrently executing jobs. Otherwise one risks seeing a delay, but not understanding why :).

The job subsystem limits are reported as max_worker_thread values, returned by

use msdb

exec sp_enum_sqlagent_subsystems