When using timeout at high concurrency, I get a different result than I expected.

[EmptyDreams]

Reproduction

I'm using the following code to make a large number of GET requests to a URL, for convenience I'm repeating the request to a URL and I've found that when the number of requests exceeds a certain value it causes most of the requests to timeout.

When I set count to 100 there is no error output from the console, but when I set count to a larger value most of the requests return a timeout error, for example if I set count to 500 over 400 of them all timeout.

const fetch = require('node-fetch')

const url = 'https://cdn.jsdelivr.net/npm/twikoo@1.6.17/dist/twikoo.all.min.js'
const list = []
const count = 500
let errCount = 0

for (let i = 0; i !== count; ++i) {
    const startTime = new Date().getTime()
    const id = i
    const task = fetch(url, {
        timeout: 5000
    }).catch(err => {
        ++errCount
        const endTime = new Date().getTime()
        console.error(`fail ${id}: use time = ${endTime - startTime}, message = ${err}`)
    })
    list.push(task)
}

Promise.all(list).then(() => {
    console.log(`finish errCount = ${errCount}`)
})

Expected behavior

I suspect this is because node-fetch provides a queuing sequence to cope with high concurrency, but the timeout timer still starts when a request is queued.

If I'm right, would it be possible to provide an option to allow the timeout to start when the request actually starts? Or is it possible to provide an interface that allows the user to customise this behaviour?

Your Environment

software	version
node-fetch	2.6.12
node	18.12.1
npm	9.8.1
Operating System	windows 10