Oded's description sounds like the consequences from changes like these:https://adblockplus.org/development-bui ... ox-startuphttps://issues.adblockplus.org/ticket/117
ABP 2.6 became available via AMO today and I did some testing with Firefox 28. Using an 8 yo single core desktop and a 2.5 yo dual core notebook (7200 RPM drive). Same Firefox and extension configs (moderately heavy protection) on both. Internet connection via cable testing around 29.1 Mbps down, 5.85 Mbps up, 12ms ping with closest test server. ABP 2.5.1 worked fine during the startup window and blocked what it should across many sites. ABP 2.6 failed to block things it should on every site I tested.
I did some more explicit testing against a lightweight localhost server using custom test pages. Under my test conditions, all of which included clearing history between each test as that is what I set Firefox to do when it closes, ABP 2.6 on the old box had a roughly 4.7 second window of vulnerability and ABP 2.6 on the newer box had a roughly 1.4 second window of vulnerability. Which may not sound like much, but that was enough to assure that ABP 2.6 failed to do its job each and every time it started up to load a page with blockable items. Some days I start it once and leave it running. Some days I probably start it three or four dozen times. Most days somewhere in between . I think the consequences could add up to very many undesired exposures over time.
I find the "No, this won't be optional" comment worrisome, as it suggests ABP will have this type of problem in all future versions. I have some generally faster systems but they are running strongly encrypted storage systems and one is loading things over a network as well. So I'm not very hopeful about those being more reliable.