I completely agree that teams need to limit the amount of “spikes” in an iteration – I’ve seen iterations with no output at all because of “spike abuse”.

Another thing you have to keep an eye on are bug fixing tasks. I’ve seen teams treating them as a separate backlog and that usually leads to either an infinite bug backlog (because of the usual lower priority assigned to them) or developers only working on bug fixes because they seem simpler to do, causing massive delay on the feature backlog.