Not op, but choosing async vs sync replication is typically done around guarantees. Async (the default mode) is generally preferred in most cases as it adds relatively small overhead. The problem with it, is that if something happens to master, the standby node might be behind (i.e. some data might be lost).

Sync on the other hand won't return from the transaction until the change is replicated.

For time series, generally async is acceptable, and even for regular data most people might also be fine with async (that's why it is the default). Sync is used if you absolutely aren't allowing any data loss. But because it waits for all nodes to confirm they wrote data to the disk writing will always be slower.

The problem is that if you use physical replication it applies to all data in the database (the logical resource not the physical server), so my guess is that some of the non-time series data has higher durability requirements.