Two things that I underestimated.
1. The complexity and length of the draft SAP. I thought it would be less than ten pages, but I was very wrong, apparently.
2. The ability of NWBO to process a continuing scrub that will, by all outward signs, be able to include all data gathered right up to the time of data lock. I thought their data would be frozen back to the end of 2018, but I was apparently wrong.