I thought about suggesting that.
in reply to Re^6: hash collision DOS
in thread hash collision DOS
The problem is that you get guaranteed worst-case behaviour at the cost of extra code along what should be a fast path, and this is going constantly be executed. I suspect that you'd find a measureable slowdown in perl. Which is why I personally prefer the solution that was chosen, which is to vary the hashing function so that its performance characteristics are unaltered but there is no predictable DOS target to go against.
However don't let my back-of-the-envelope guess deter you from trying to implement that solution. It may be that there is no significant performance penalty, in which case the improved worst-case performance would be very worthwhile.