Initial system start performance

[elmeister]

Switching from @std/esm to esm increased the time between npm run start to Started on port 8080 in our system from about 10 seconds with @std/esm to about 30 seconds with esm.
Is there any general advice - if we can try to change some options, or enable debug to track why this may happen? Maybe there are some particular bottleneck qualities that we could pay attention to?

[elmeister]

@jdalton, we appreciate what you do a lot, but we really need help.
Any chance someone will look into this issue some day? When we need to debug our code, it makes a huge difference between 10sec and 30sec system start up time. And now, without the support of newer versions of Node in @std/esm and without support of older versions by AWS, we are in a bad position.

[thomvaill]

I have exactly the same issue in this project: https://github.com/thomvaill/log4brains
We use esm for our CLI and it adds a constant startup time of ~3s (which is a lot for a CLI).
Even if the cache seems to be written, it does not seem the improve performance.
Actually, I see no difference when ESM_DISABLE_CACHE is set and when it's not.
Maybe this is a cache issue, but I don't know how to debug this.
Could you help?