Dynamic platforms where components can be loaded at runtime can introduce risks to applications stability if components are not previously known before deployment. It may be needed anyway to execute such untrustworthy code, even if it is not malicious. The OSGi platform presents such a scenario where components can be installed, started, stopped, updated or uninstalled during application execution. In this paper we describe how we introduced dependability and monitoring as crosscutting concerns in the OSGi platform for improving applications resistance to such risks. These concerns crosscut different software layers which are well defined in the OSGi specification but scattered over different places in the OSGi API. We also created a level of indirection by representing software layers as aspects, enhancing the API's modularity as well as reuse by avoiding redundant pointcut definitions. The dependability aspects helped us validating the layer aspect abstraction reuse. Since the aspects targeted the OSGi API, it was possible to weave our solution into distinct versions of three different OSGi implementations, namely Apache Felix, Equinox and Knopflerfish. We validate our approach on all of the woven platforms in a simulation of an RFID and sensor-based application that uses untrustworthy components.