Ensuring TTI measurements are accurate

[fabiospampinato]

Looking at the top results for the TTI metric I see some super strange numbers:

1. First of all what unit do these numbers have, milliseconds I guess?
2. The vanilla implementation doesn't seem minified, arguably it should be minified to be a more accurate baseline to compare against?
3. All these numbers seem strangely close, could it be that we are actually mostly measuring how long it takes to download/parse/execute the CSS? If so IMO that muddies the waters significantly and loading the CSS files should be disabled when measuring the TTI metric.

[krausest]

(Sorry, my free time is currently a bit limited, so I can't put as much effort into this as I'd like to do).

1. consistently interactive is the "interactive" audit from lighthouse. It returns the following JSON:

{
    "id": "interactive",
    "title": "Time to Interactive",
    "description": "Time to Interactive is the amount of time it takes for the page to become fully interactive. [Learn more about the Time to Interactive metric](https://developer.chrome.com/docs/lighthouse/performance/interactive/).",
    "score": 1,
    "scoreDisplayMode": "numeric",
    "numericValue": 1873.368,
    "numericUnit": "millisecond",
    "displayValue": "1.9 s"
}

2. Here's a comparison for vanillajs and vanillajs w/rollup + terser soon! (It used to be webpacked, but at some point in time we removed bundling for vanillajs 😄 )

If you can propose more sensible lighthouse audits we could switch to those...

[fabiospampinato]

I'd propose ignoring the CSS for that test, which seems the main problem causing those numbers to be all that close together.