Deleting the whole profile (rm -rf ~/.config/chromium) and opening the URL still shows Number of visits > 1.

Thanks for the report!

As far as I know, we're using an in-memory cache implementation in Incognito, so this definitely shouldn't happen. Nevertheless, the bug reproduces consistently. Upping priority to 1.

I played with the website and observed the following:

a) Regular mode: If I type something into the text field and reload the page, the text stays.
b) Incognito mode: Same behavior.
c) Regular->Incognito: If I type something into the text field and then open the page in Incognito, the text stays (this is the bug), but after the next reload, it's gone.
d) Regular->delete->Incognito: If I type something into the text field, then delete cache, then open Incognito, the text is still there.

I think that observations c) and d) imply that this can't be an issue with the HTTP cache being shared between regular mode and Incognito. If it was, reloading in c) would keep showing the text, and the deletion in d) would conversely result in not showing it at all.

+component:Cache

Cache folks, do you have an idea what could be going on here?

+component:Blink>Loader

Shot in the dark - is it possible that we're not caching an ETag in the HTTP cache, but somehow an entire rendered page somewhere else and reuse it once in Incognito?

I looked into the code of this page and it seems like they are cheating:

else { // No etag was sent. We need to generate one. Normally you would derive this from randomness.
		$etag = substr(sha1($secret . sha1($_SERVER["REMOTE_ADDR"]) . sha1($_SERVER["HTTP_USER_AGENT"])), 0, 18);
	}
(https://github.com/lucb1e/cookielesscookies/blob/master/index.php#L15)

Chrome correctly deletes all etags when cache is deleted or incognito mode is used but the site just generates the sessionid from your ip adress and user-agent. This website shows that fingerprinting works but there is no bug in Chrome. It can also be observed in devtool that we send no etag after cache deletion.

There is a discussion on the bug tracker about this issue: https://github.com/lucb1e/cookielesscookies/issues/18