> If a work year consists of 300 days, that would mean almost 18,000 man-years, which, spread over 20 years, implies a workforce of about 900 men.
While the analysis is interesting, I think the 300 working days a year is too long. Likely, work on these types of public projects was mostly done during the flood season, which was around 4 months (120 days [0]) so maybe 100 work days at best, so the calculations for required workers should probably be multiplied by 3.