What Molina cannot do
What it cannot run
Software that only runs on Windows. Software that needs a GPU or special hardware. Software with no container image.
What I cannot promise
I cannot promise to find a bug. A mature library might have nothing to find, and then you get a report of what was explored. A clean report means what was explored held up, and nothing more.
Replay on your software
Replay matched exactly in five of my runs: an Akka.NET cluster, a .NET program with racing threads, Postgres, NATS and Akka's own test. Replay is not yet measured on your software until I run it. Some software fits exact mode, and some runs in screening mode, which compares what was found and compares no bytes. Exact mode refuses some setups today, such as a disk attached to one of several services, and stopping and restarting a service.
Where the power is cut
The power is cut at a point in your workload that I choose: a numbered disk request. I have not built a way to cut the power at a chosen moment of the virtual clock.
The Postgres count
The number of saved writes that Postgres lost moves a little between runs, so the count is not a replay result. Postgres's own threads and timers are not controlled.
Running it yourself
You cannot run Molina yourself yet. There is no installer, and I run every failure again on request.
The recording
You read the recording and check it by its hashes. It is a folder of plain files and one web page that opens in a browser, made with molina bundle page.
Whose runs these are
Every result is from my own runs on one machine, and each says what it does not show.