I wasn't aware of the different handling of "script-inserted" scripts and I must admit I misunderstood the async /defer attributes. I was hoping to get on-demand parsing out of it; control over the point in time the browser starts parsing a previously downloaded script.
The idea is to test if we can get better perceived performance by loading scripts during the apps idle times, but defer parsing and executing it until user interaction necessitates it.
If my assumption is correct, one could more easily balance between startup time and perceived performance on low-bandwidth ,high-latency networks.