Quick questions about Nova.AiLab

Hi,

I was checking out Nova.AiLab and really liked the idea of testing branches externally like this.

Had a few quick questions I was curious about:

How do you handle environment differences when building via NovaRepo? Could that affect results?
Are runs (like match --repeat) fully deterministic with the same seed?
In practice, how do you decide a metric improvement is actually meaningful?
Do you rely more on the dashboard or the markdown reports in your workflow?

Really cool project overall—curious to hear your thoughts