Is a virtual thread faster than a platform thread?

A virtual thread runs ordinary bytecode on an ordinary platform thread, its carrier. Nothing about the code itself gets faster, and one request takes as long as it did before.

What changes is the price of waiting. A platform thread is an OS thread with a stack reserved up front, so a server can afford a few thousand of them. A virtual thread is a small heap object; when it blocks on I/O or a lock it unmounts and frees the carrier for another one. A million of them waiting on sockets is fine.

So the win is throughput for I/O-bound work: with many requests each spending most of their time waiting, you can have one thread per request instead of a pool that runs out. For CPU-bound work there is no win at all: the carriers are as many as the cores, and the cores were already busy.

A good answer ends with the caveat: more concurrent requests means more pressure on whatever they wait for. The database pool is still ten connections, so bound it with a Semaphore.