While working on the performance improvement using the wtperf workload update-only-btree.wtperf, even when I disabled storing validity window on the disk (hack mentioned in WT-6082), I saw there is activity related to the history store. We are writing something into the history store and even truncating records because of the mixed timestamps.

Though it is not a fair comparison, in contrast, there was no lookaside activity before durable history.

I could disable the function that writes history and the workload was not affected. It will be good to put some effort into investigating why we are writing to the history store when timestamps are not used at all.

I saw no impact on performance, so this doesn't need to be scheduled immediately.