Patchwork provides the generate_perf_data management command to create synthetic projects, submitters, series, patches and replies for local performance testing. You can use this with a disposable database, configured like the database backend of the instance you want to measure. You can use the command like so:

python manage.py migrate
python manage.py loaddata default_states
python manage.py generate_perf_data --projects 12 --patches-per-project 100000

The defaults create 1.2 million patches in batches of 1,000, with up to ten patches per series and 100 shared submitters. Each patch has between 1 and 10 replies (roughly 6.6 million replies with the defaults), with randomly chosen submitters and dates after the patch. Reply counts and contents are reproducible with --seed. Patch dates span a year. All patches contain the same small diff: this is a query-performance dataset, not a realistic simulation of database storage requirements or mail parsing.

Generated projects have names such as perf-1. The command refuses to reuse existing project names or generated submitter addresses. Use a different prefix for another dataset, or recreate the disposable database. It never deletes data. Batches are committed independently, so an interrupted run leaves partial data. Save hooks and signals are bypassed: events, tags, checks and cover letters are not generated.