PostgreSQL dblink's way of manipulation with connections

I have created the named dblink connection, then used db A to call
simultaneously the same procedure twice on db B. I have noticed that
the there is one open connection all the time, and there is second
which appear and disappear continually. Is there a way to use
connection pooling, so that one call could take up one connection and
free it once it has done calculating? Or does dblink use completely
other way of manipulating connections?