When people tell me they've learned from experience, I tell them the trick is to learn from other people's experience. – Warren Buffett
When you start a company, you get three innovation tokens. You can spend them however you want – adopting a new framework, implementing a unique interview process, adding non-standard terms to a term sheet, 0r designing your own database all count as an innovation token. The standard advice has been to choose boring technology. And often, that's not a bad bet (see Getting to Market with Rails).
Yet, innovative and boring technologies are both dangerous. Boring technology that misses stepwise improvements is spending an innovation token, as it needs to catch up to the status quo (see Engineering Against the Grain for some examples).
Innovation tokens are often more a function of the team's current and future skill set than old vs. new. It can also be a function of the current state – see the problems with Toolchain Sprawl.
A different way to think about the risk analysis with innovation tokens is through comparative and absolute advantage. For example, your engineering team might be best-in-class. Yet, having them work on compilers is unlikely to move the needle for your enterprise SaaS startup (well, there's always an exception). Having an absolute advantage across many buckets doesn't give you a comparative advantage (and creates a higher opportunity cost for anything other than your most efficient bucket).