Simply put: We can't.
Throwing aside all the contract and interoperability issues she raises, moving all your crap from one system to another is *hard*. Changing your workflow is *hard*. Retraining your users is *super hard*.
I've often commented to people that if all the time and energy that has been put into trying to "put lipsick on a pig" with OPACs had instead been spent creating converters to easily move data between the systems the major players sell, we'd have a hell of a lot more competition and, one can assume, better products.
Is open source the answer? I'm not sure it is. I *am* sure that there's a place in the market for new players, esp. among the smaller libraries who are (a) hungrier, (b) less beaurocracy-encrusted, and (c) more likely to look at hosted solutions. I'm not sure I'm the guy to fill that niche, but boy oh boy do I hope some folks start to step up.