Why slave_parallel_workers affects TPM

I ran a few HammerDB benchmarks and mess with the configurations. And I notice different values of slave_parallel_workers have a considerable impact on TPM.

As far as I understand, slave_parallel_workers only affects how quickly the transactions are being applied from the relay log. Yet, all the possible HA/DR setups, including async replication, semi-sync replication, and group replication will not wait for the transactions to be applied on the secondary node(s) before committing. To my understanding, slave_parallel_workers should only affect replication lag and have zero impact on TPM while this is clearly not the case.

Am I missing something?

Edit: To clear any confusion, I expect the value of slave_parallel_workers will have no impact on TPM. While the reality is the higher the value of slave_parallel_workers is, the higher the TPM is. So I’m hoping someone would be able to explain the reason for that.