Before processing a new task, VBR detects involved backup infrastructure components (backup proxies and repositories). When a new job starts, VBR analyzes the list of VMs that the job includes, and creates a separate task for each disk of every VM to be processed. Tasks in the job can be processed in parallel
Task limiting is performed by the Veeam Backup Service that is aware of all backup proxies and backup repositories connected to it, and settings specified for these backup proxies and repositories. When a job starts, it informs the Veeam Backup service about its task list and polls the service about allocated resources to its tasks at a 10 second interval after that. Before a new task targeted at a specific backup proxy or repository starts, the Veeam Backup Service checks the current workload (the number of tasks currently working with the proxy or repository) and the number of allowed tasks per this component. If this number is exceeded, a new task will not start until one of the currently running tasks is finished.
Note, that each task requires a single CPU core. Does your proxy have at least 5 CPU cores? How many jobs are you running in parallel? Are there several VMs/disks?