If it ain't broke but obselete, should we upgrade it?

My MBA class and I discussed the case of a successful retailer whose business model was heavily dependent on their IT infrastrcuture, and whose IT infrastructure was DOS-based. The class initially seemed unanimously in favor of upgrading the system to a more modern platform. However, a parallel analysis led us to conclude that the existing system supported the retailer's business model and information needs very well. There were no obvious immediate deficiencies, their IT was well aligned with their business strategy, their processes were carefully engineered, and their employees were trained and empowered. Besides, their key challenges to scaling their business were not IT-related, but had to do with operations and human resources. 

The discussion ended unresolved. A good summary of what was left open might be: when a system is technologically "very old" but functions in perfect alignment with one's business model, does it make sense to rip it out and replace it towards being technologically current, or reducing vendor dependence, or to create future (but non-obvious) application development options? Or should IT investment strategy always be driven by forseeable business needs, alignment with current business strategy and the philosophy of "if it ain't broke and does what you need well, don't fix it"?