You can submit several tons of jobs to make up a particular batch run. If hey are not TYPRUN=HOLD, some will start executing and some will fail in the initiation as datasets are not yet created, etc. They will only run in the order you want by wild co-incidence.
BUT, If you do TYPRUN=HOLD and have a scheduler, then (assuming scheduler set up to operate like this) all would be OK. The scheduler would do the relasing when the prependencies are satisfied.