I attached the WebIDE debugger (F8) to a Firefox Beta for Android instance (Firefox 61, buildID 20180514150347), inspected the main process and switched the context to an extension (with the tabs permission so I can see the URL, e.g. uBlock). In the browser all tabs except for the last one were zombified/"discarded" (I guess that you can trigger this by causing memory pressure, e.g. by opening another app that uses lots of memory). browser.tabs.query({ discarded: true, }).then(tabs => console.log(tabs[0].id)); Then I triggered a navigation, e.g. browser.tabs.update(4, { url: 'https://example.com', }); I can see that the tab has navigated. However, when I query the tab info, then the tab is still in the "discarded" state: browser.tabs.get(4).then(console.log); // { ... discarded: true ..., url: "https://example.com/" ...} When I switch to the tab, the original URL is restored. What I expected is: - After navigating by tabs.update, tab.discarded should be false. - After tapping on the navigated tab, the tab should not be "restored", but stay at the changed location. The implementation of tabs.update calls nativeTab.browser.loadURI : https://searchfox.org/mozilla-central/rev/83a923ef7a3b95a516f240a6810c20664b1e0ac9/mobile/android/components/extensions/ext-tabs.js#306,318 The zombie restoration logic scattered over: - https://searchfox.org/mozilla-central/rev/83a923ef7a3b95a516f240a6810c20664b1e0ac9/mobile/android/chrome/content/browser.js#3922-3963 - https://searchfox.org/mozilla-central/rev/83a923ef7a3b95a516f240a6810c20664b1e0ac9/mobile/android/components/SessionStore.js#846-859,1462-1480 (Search for "pending" (including the quotes or __SS to find relevant code). It seems that the tab management implementation does not account for navigating a zombified tab.

Conversely, tabs.update() shouldn't be stomping over the session history of a discarded tab, though. In principle, the tab needs to be restored first before we can load a new URL the normal way, although a naive implementation would trigger an additional, unneeded load of the previously loaded URL, which isn't ideal. I wonder what Desktop does here?