Is DBD::Proxy setting up a single connection and then passing it to each test script, as asked by the OP (thinking along the lines of passing a file handle between various scripts), or is it setting up a meta-connection, where a connection to a DBI::Proxy connection is talking to another DBI connection through, well, a proxy?
This approach is different than what I was considering, and does make some good sense. An approach that I had not considered!
Update: Reading through the documentation for DBI::Proxy, I am not certain that it meets the needs of the OP's third requirement - connecting once to the database and being able to run all of the test scripts. Perhaps it is just my unfamiliarity with the Coro method of doing things. Would you have a more complete example that implements something along what the OP was asking?