measure sync time in a non-instrumented run
looking at this graph: https://benchmarks.leap.se/test-dashboard_test_encrypt_raw_10k.html
and considering that the threshold is related to beginning of cpu and mem measurement, I find that it would be better to have some benchmark runs for sync (encrypt + upload, download + decrypt) in a run that DOES NOT measure cpu and memory in the same run (ie, fix one of the metrics and stop other instrumentation).
The goal would be to have time measurement with less dispersion.