Actual latency sensitive apps can always use SOCK_RAW and implement their own TCP. In fact, for serious low latency you need to bypass the entire kernel stack too, like DPDK.

“I know, instead of using TCP_NODELAY we can just roll our own ad-hoc reimplementation of TCP!”

For serious though. TCP has a handshake to make a connection. UDP doesn’t and if you’re going for low-latency, you probably don’t want retransmission anyway, since if you miss the packet it’s too late to do anything about it. If you need the guarantees from TCP, you probably aren’t actually solving a low-latency problem.

Low latency SQL transactions? Not microseconds, but a ms or two is doable unless you have transient connections or Nagle + delayed acks.