Errrrrrrrr - small puncture. Repeating crossbows are *not* awesome weapons. Neither are repeating ballistae. The mechanics of the 'automatic ballista' as reconstructed by various archeologists do not lent themselves to use in a crossbow, anyway, because they would require three hands on the part of the operator, but that could have been solved.
More to the point, the Romans had both handheld crossbows with tension systems (probably composite, though no bows survive) and nut trigger releases and heavy but man-portable 'siege' crossbows with torsion systems and claw trigger releases. Both would by our standards be considered more 'advanced' than the 11th-12th century selfbow, groove-release crossbows that many military historians believe to have been such a devastating novelty. (The handheld tension catapult, while long theorised, was widely dismissed as a figment of the imagination until they excavated one in Germany last year...)
The problem with repeasting systems for crossbows is that they don't really produce any real bonus. The reason for that is that crossbows are fundamentally different from guns in where they get their energy. A gun gets its power from chemicasl energy stored in the ccharge. If you can cut short the process of inserting the charge into the weapon, you have cut down the time bertween firing considerably because that process is the most time-consuming part. Catapults and bows, on the other hand, get their energy by tensioning a system, and irf you ciuts down the time of placing the arrow into the tensioned system, the most time-consuming part of the operation - tensioning - remains unaffected. You can work around that by reduciong the tension, to speed up reloading time - the Chinese repeating crossbows do that - but of course that gives you less power. Roman repeating designs have been reconstructed, but never yet shot at anywhere near the possible power, so we have no idea how they would perform 'loaded for bear'.
Finally, the Late Roman armies most likely did utilise crossbow units (we do not know whether they bused tension or torsion designs). Apparently, they performed well enough to stay in service, but not spectacularly better than archers, who continued to outnumber them. And given that they trained all recruits in archery at that time - a time-consuming process - they must have figured that was the better move. I can see why - bows get better energy conversion, so archers can achieve vastly higher rates of fire with killing projectiles than crossbowmen can. Crossbows can get more energy in (but trade for even slower rate of fire), can be more accurate, and above all are easier to use, but none of this strikes me as a concern in the Late Roman army's situation.