We had fewer hands and less time. The real problems were RAN-related and infrastructure-related, mostly due to poor management of the preparations on my end. I had a lot of unexpected distractions in July that screwed my schedule. I apologize.
On Sunday, the main problem with SMS delivery was not that there were not enough delivery attempts, but that so few of the delivery attempts were successful, probably because by Sunday afternoon, 2 of our 4 BTS units were offline and the remaining 2 had shifted into automatic power reduction to try to protect themselves from the high registration loads. Having fewer tower sites didn't just mean less coverage; it also meant less total RAN capacity, which concentrated the registration loads into fewer channels and meant less excess capacity to absorb a BTS failure.
Should we do it again? Yes, but only if we can stage properly at least two months ahead of time. This could be done in our SF test network. Coordination with the City of SF has been slow, but is finally moving.
It is significant to note, we had significantly fewer technical hands than we did last year. Our fewer towers was not lack of effort, rather time.