If we detect an abnormally large number of reviews that will be due in the near future, we'll have you perform some reviews earlier than you optimally need to.

A common issue with spaced-repetition tools is inconsistent daily requirements of "reviews due". One day you might have 5 minutes of reviews to complete, but the next day: 25 minutes.

This happens naturally and isn't a failure of the algorithm; your performance with each word is tracked separately, so having some days with more reviews due than usual is to be expected. 

However, a wildly inconsistent number of reviews due each day can be overwhelming, and is one reason why some people stop using spaced repetition.

Which is a shame, because the algorithm hasn't stopped working.

In fact, it is during these times that continuing to trust in the algorithm's scheduling makes the most sense: you'll refresh your memories of those words most in need of refresh first, before refreshing memories that will probably last a little longer on their own anyway.

While this is less "optimal" from a purely mathematical standpoint -- we aren't robots. Reducing the psychological burden of a huge pile of reviews due will likely keep you in the game, which could be net positive in the long run.