The abstractions are super leaky. Infrastructure fundamentals are consequently significant in all applications. Failure or refusal to understand and inwardly digest this leads inexorably to bad software, or more precisely, bad systems. Web developers that do not know TCP/IP and do not wish to are a consequent burden to everyone else; a fountain of horrors arises, from latency issues to security blunders, when mechanical sympathy is disregarded.

I still disagree. As a web dev, there is absolutely nothing you can do to affect the networking stack. This is all completely buried in web servers and the like. Most CDNs are complete SaaS tools where you can't anything beyond the highest-level. Everything devs do to optimize performance is at the application layer. You can observe things like latency or number of connections and change your code or infra, but you're absolutely never going to inspect a packet. Unless you're actually developing those web servers.

I have countless times inspected packets on the wire when writing web applications, and there is tons we do to ensure service behaviour aligns with the networking stack.

I have also inspected traffic when investigating/writing patches for server code or the kernel, but that’s a completely separate matter.

What’s more, most nontrivial application code causes additional network activity, from DNS queries to database connections, calls to external APIs et cetera. The all-the-world-is-HTTP fallacy is how wheels get reinvented, but badly, as with websockets or JWTs.