So far, we supported events which had up to 2^16 = 65k clusters.
In the processing of 2024 data, some events were encountered with many, many more clusters.
While these are pathological events that (I think) we want to reject with a GEC, there was actually not a very strict reason to have the 65k cluster limit. This shifts that limit to the maximum possible size of an std::vector, instead.
