While there is no denying the fact that the Apple store has a developer-store trust model which has distinct advantages when it comes to dealing with malicious apps, the more trusting approach adopted for the Android market is a major asset for mobile development teams which follow an agile approach in their work. The main reason for this is the minimal time between publishing a new version of an app and the app becoming available to end users. In contrast, the lag between submission and update for apps on iOS is measured in months rather than minutes.
In GeoSciTeach we used this feature to pursue a rapid release schedule, which in fact accelerated to several updates per day near major dates such as user experience workshops and other fieldwork tasks. Using the statistics feature of the market, we were also able to monitor our users, probe those less quick to respond to changes, and quickly identify and fix critical bugs that led to the app aborting. Although such updating is probably somewhat less dynamic compared to the web, it is nevertheless a hugely useful feature and the system overall is dependable across devices and platforms.
On the latter point, we have not experienced any difficulties in using the apps across a variety of devices (Google ADP1, HTC Desire, HTC Desire HD, HTC Wildfire, Motorola Milestone 2, ZTE Blade and others provided by the volunteer testers) and the app scaled nicely without any modifications on the Samsung Galaxy Tab 10.1.
However, Android has many more goodies for experimentation with users. Though the tight integration of Google Analytics it is possible to record and analyse every user action, so that the effects of every modification can be evaluated though meaningful metrics.
We will be looking at this for the final version of the app, which will hopefully lead us towards beta testing. Although the combination of market and analytics seems to fall just short of allowing full A/B testing, it should still be a major addition to our toolset for testing.