addyosmani
commented
4 years ago I'm +1 to us expanding our language around why a metric may have an inflated value, similar to how we speculate tab backgrounding may have been a cause. Is there a particular condition you think we should be using to signal showing this message?
On the lower end of the complexity spectrum, we could set an upper bound (e.g LCP > 1000s) after which this reasoning is given. On the other end, we could do some time book-keeping about the length of time a tab has been open. I'm personally leaning towards something simple... but am open to ideas :)
LCP should stop updating after user input, but in a long-lived tab I found that it's still periodically updating. For context: https://twitter.com/rick_viscomi/status/1278567773657092097
Should we add a note to LCP that the value is inflated, similar to when the tab was loaded in the background?
I'm also interested to find out how it got this high in the first place and if this is considered a bug.