I'm not a qualified networking engineer, but we had some US users report outages which we investigated from US-based servers and noticed the normal routing was screwy (in the sense that the normal direct routes weren't being taken so latency was high). These incidents lasted under ten minutes and I wrote them off as being peering related incidents as they had no effect on European traffic.

I accept minor things like this as part of the price for hosting in Europe, and have certainly experienced the same issues going the other direction (i.e. me reaching US-based services) more often.