2) VCs are often the vector by which this all happens. They ask their portfolio companies to pull together the info for their employees, presumably submit it into the companies aggregating everything, and then the startup gets a copy of the recent data.
3) Even done the old way (Excel), the data was incredibly detailed. You can slice and dice by startup stage (series A vs series B vs seed), employee count, region, sector, etc to determine if you're paying market rates or not. This is particularly useful for growing startups, where the founders have no idea what to pay, say, a VP Marketing at their pre-revenue mobile gaming startup in Helsinki.
4) Obviously whether or not this is bad for employees themselves is debatable, but I think people are missing the point that these surveys are ALWAYS skewed UPWARDS due to the much higher volume of data from the large tech companies (because they have far more employees and tend to be offering significantly higher comp). So in practice, the impact is likely to RAISE wages at earlier stage startups who are competing for the same talent as later stage tech.