How does lighthouse network activity influence LHS score?

[sdmcraft]

I want to understand how does lighthouse wait for network activity to subside.
For context, I have 2 test pages:

• fast.html (renders some content, waits for 3 seconds, renders some more content)
• slow.html (renders some content, waits for 3 seconds, renders some more content, during the wait, keeps doing GET requests every 500ms)

What I observe here is that fast.html consistently get 100 LHS while the slow.html has its LCP and SI in 90s. The LCP for slow.html considers the content rendered delayed also as part of LCP while the fast.html's LCP does. not have the delayed text as part of the LCP.

I have feeling that the constant network activity prolongs the LHS audit, making it to take the content rendered in a delayed way into consideration unlike fast.html.

I am using https://pagespeed.web.dev/ for running the tests and here are the audit results:

• fast.html's LHS audit
• slow.html's LHS audit

Looking for deeper insights as to what happening here.