We've just made some changes to our backend SRS algorithm in response to suggestions that people aren't getting enough exposure to the vocab items before the words start disappearing into the SRS void.
With this in mind, we're trying an experiment that will require words to be tested correctly 3 times before the normal SRS mechanisms kick into effect. This isn't set in stone, but for heavy users of the review or flashcards, I'd be curious how and if you like or notice the difference. It basically means more immediate up-front testing, but hopefully not enough that everyone feels we're beating them over the head with words they already know.
If you notice any issues or have any feedback (positive or negative), please let us know. This is the sort of thing where what we hear from you is likely to determine system defaults.