Ok. Let's assume that all columns had the strength to carry 3 times their normal load. Let's assume all columns were in unblemished state. If a moving top block now falls onto all columns perfectly and absolutely simultaneously, then we would expect to see an (average) deceleration of -2g (for a very brief momentm, until columns are overloaded).
However if only half the columns impact, and the other half an instant later, we'd see only -1g, for twice as long.
But our assumptions are wrong, on several counts:
- Column-on-column impact would not be perfect, and not happen at all in several places
- Not all columns would still have their full strength, as thos impacted early on were already stressed by heat, sagging floors, etc.
- Impacts are not as vertical as the static loads the columns are designed for
- The factor 3 is possibly an overstatement. NIST assumes 1.667, IIRC.
If we correct all assumptions, we will probably find that the maximum expected deceleration would be much closer to 0 than -1g. Which increases the probability that your low sampling rate will miss that peak.