In the "Effectively Troubleshooting Latency and Failover of Always On" by Sourabh Agarwal At PASS on Friday, November 3, 2017 3:30" about 30 minutes in he talks about how Synchronous mode commits are quicker, as the hardening is just about getting the t-logs to disk on the secondary. The Redo is an async operation outside of the AlwaysOn commit process.
I may be missing something, but it seems like this "could" mean that the primary OLTP will not be impacted by long running (and blocking) queries on the secondary. I envision that the logs just build in size on the secondary until the blocking report query completes, and then applies all the changes, that have occured on the primary and been hardened to disk on the secondary.
Related Will running a large query on a secondary database in an availability group affect transaction performance in the primary database? In this question the answwer by Kin suggests the answer does not address 2016+
Does SQL Server 2016 eliminate poor primary performance when reports take locks on seconday?
For this question: Assume multiple disks, with data (MDF) and logs (LDF) on seperate drives.