main

The load generator of the end-to-end benchmark (test/benchmarks/README.md):

• seed: opens the accounts, each answered once its snapshot holds it;

• check: one operation of every scenario, each checked as a run checks it;

• prime: an untimed pass that touches every seeded account once, so a run on a large data set measures warm caches rather than filling them;

• settle: after the prime, waits until every node is quiet again, so the warm-up does not start on its backlog;

• run: one open-loop run of a scenario at a fixed rate, its results and resources written as JSON and Markdown;

• saturate: the closed-loop saturation finder.

Options are --name=value; see usage.