LW has a continuous onslaught of crawlers that will consume near-infinite resources if allowed (moreso than other sites, because of its deep archives), so we’ve already been through a bunch of iteration cycles on rate-limits and firewall rules, and we kept our existing firewall (WAF) in place. When stuff does slip through, while it’s true that Vercel will autoscale more aggressively than our old setup, our old setup did also have autoscaling. It can’t scale to too large a multiple of our normal size, before some parts of our setup that don’t auto-scale (our postgres db) fall over and we get paged.
LW has a continuous onslaught of crawlers that will consume near-infinite resources if allowed (moreso than other sites, because of its deep archives), so we’ve already been through a bunch of iteration cycles on rate-limits and firewall rules, and we kept our existing firewall (WAF) in place. When stuff does slip through, while it’s true that Vercel will autoscale more aggressively than our old setup, our old setup did also have autoscaling. It can’t scale to too large a multiple of our normal size, before some parts of our setup that don’t auto-scale (our postgres db) fall over and we get paged.