Typically, 2 NEMA23s would be much more than a single A4988 could drive, quite possibly more than a DRV8825 could handle. They tend to draw 1-3+ amps per phase each, depending on the winding and depth. That said, even if you had a low amperage NEMA23, you might have a problem running fast, with the inductance of that nema, since it would probably be a higher voltage rated NEMA23 to facilitate low amperage.
You have a couple options.
(1) wiring the NEMA23s to two different drivers. You probably need to set it to “FULL step” if you do this, since microstepping might be a tiny bit off between the drivers (I dont know fully, but I understand you can’t assume the two to work exactly the same when driven by two drivers in microstepping.) The good news is you can wire the EN/STEP/DIR to both drivers in parallel, so the code doesnt need to change.
(2) Running an external driver will work too, as you mentioned it, but you want to make sure your external driver can handle 5V singles, since the RAMPS is a 5V device.
(2)a), I don’t know if you have access to decent stepper drivers, but the ones better than DRV8825 tend to cost a bit of $$. There is a common Toshiba 3.5A-4A driver that is cheap on ebay, but do your review searches because I understand they are often poorly made and can blow up pretty easily.
(3) if your code, and I believe RAMPS is ok, does not have a 2uS pause on the signal, it may have a problem with some drivers like the DRV8825 which requires a >2uS pause (the a4988 is 1us). This usually isn’t a problem though, so I dont think you need to worry.
Brandon is right that NEMA23 isn’t enough information, but “chances are” the a4988 is not sufficient. You suggest wiring an external driver…and, yeah, if it follows the step/dir protocol, and can tolerate 5V, you should be in business.