Is that actually what they need?
It sounds like they need capacity and they're unlikely to be able to get that by rewriting code at this point.
Legacy COBOL programs tend to be tied to legacy databases and legacy hardware so it's unlikely blue-sky thinking will involve Azure any more than Cobalt.
They probably need first to consider solutions like replicating the application on any spare hardware they can find and dividing the workload between them in such a way that the data can be consistently reconstructed later (such as by first letter of surname).