Complete change freeze so you only need a minimum number of developers and ops people around. Although it is sometimes a good time of do that big CICD pipeline rebuild.
The lawyers, HR, finance and office support are down to minimal. Same with the cleaners (a good time to repaint the office though), internal IT, cafe and other support staff.
Sales will be down to minimal since all the customers will be on holiday. Customer support might be at normal or reduced level depending on industry.
LinkedIn shuts down for 2 weeks a year. These are the best vacation weeks. Like others say, there is no email waiting for you. No meetings or decisions that you regret not being involved in, etc.
The net amount to the public may be small. The scale offered relative to the size of the companies, and within high-value niches (enterprise customers, business partners, vendors/suppliers) may be significant.
Coordinated downtime can still be an advantage. It's actually a guiding principle in CPU scheduling systems as well -- so many cycles dedicated to processing vs. overhead.
Back when "Silicon Valley" meant hardware and production lines, the practice was to shut down entirely for at least a weak between Christmas and New Years, in cases two, as 1) operating the line with short staffing increased errors and costs and 2) stopping and restarting the line incurred large losses as well of material in start-up and shut-down phases.
Better to entirely curtail ops for a couple of weeks.
That tradition had largely continued to manifest even with the offshoring of wafer, disk, and chip production, through the 1990s and aughts, though it's been in a slow fade for a long time.
Service industries operate on a different clock, of course, but a scheduled downage can still be communicated.
Then there's Burning Man, when SF-SJ palpably empties out. New Years in August.
