First of all, pay closer attention to the usage patterns. We saw increased activity for a few days before the beta signup, but never investigated in depth. We only learned that something's up when the heavy traffic started and monitoring alarms went off.

Second, get the scaling-up procedure more automated, we had everything planned, but it took too much time to get the extra infrastructure ready. Pressure and precise operation don't work well together.

Third, we uncovered a previously unknown memory leak in one of the components, which got evident during the Guild Wars 2 event and forced us to restart a subsystem a few times.

All in all it was a nice (if a bit stressful) experience, hope you guys enjoyed it a bit yourselves too :)