There's an important difference here between latency and throughput of the payment gateway. You may have four instances of a single script running, but if each script is only capable of doing one transaction at a time, then you probably aren't coming anywhere near the theoretical performance that you might obtain. This is because there is a 'lag' between the time you send a request and the arrival of a response from the gateway.
If a single script can line up a hundred requests at once, you might be able to dramatically increase your throughput. PHP is not meant to run as a multi-threaded application, but the [man]curl_multi_init[/man] function allows you to create a very large number of concurrent requests.
Here's an analogy: You currently have four 'assistants' that walk into a store to perform a credit card transaction. Each has to park their car, walk in, choose a product, checkout, and come back to their car. There's an inherent slowness to this process which cannot be sped up. However, if you hire 100 more assistants, then you can dramatically increase the number of products you buy in a given day.
Hope that makes sense.
EDIT: corrected me poor english