I was mostly wondering about arguments for run command to specify cpu/memory so it's easy experiment with them on demand.

Honestly, I've started working on this PR with this mindset and realized that it's not a good idea in terms of transparency.

Here's why: we already store VM's configuration in a file. Starting a VM with configuration other than the one described in it's configuration file is confusing in the event the user chooses to inspect the configuration file (#1 (comment)) and infer some knowledge from it. It will be misleading.

Plus it will play nicely for persistent workers. No need to clone-set-run, can just clone and run right away.

This process is fully automated, so no issue here at all.

Do you think it makes sense to include these changes here or in a separate PR?

I'd vote for a separate PR.

I'd vote for a separate PR.

SGTM. There is already #4 for that actually.

On a second though why do we store cpu and memory for a VM? What about storing only the min values and have flags like --cpu and --memory for run command with some feasible defaults like 2 CPUs and 4G of memory? Don't think it there will be a VM of M1 machine which won't comply with these requirements. Then there won't be a need in set at all and when a persone will run an existing VM it won't be too slow with the defaults.