Emergent software systems are composed, and continuously re-composed at runtime, from a large pool of small potential building blocks with the aim of responding to changes in the deployment environment. The approach assumes that each building block (such as hash table, cache, or scheduling algorithm) has a set of available variation in the selection pool, such that the most appropriate collection of variants can be composed according to the current conditions, with some objective function in mind. Populating such a pool of implementation variation, however, is not a trivial task, and existing work has examined the use of Genetic Improvement (GI) to drive this process. In existing GI-driven approaches in this space, because the building blocks being considered are very small and are improved in isolation, researchers have used a mixture of new code synthesis with traditional GI mutation/crossover operators to gain sufficient novel genetic material. To cope with the resulting search space size in a general-purpose programming language, research has suggested the use of phylogenetic analysis to help drive (rather than to explain) evolution; this examined the independent effects of both crossover and different mutation types, and was also used to demonstrate combined lineage selection and optimisation in a single run. While this use of phylogenetics offers a level of automated search guidance, by itself it remains limited to single genetic pools. In this paper we propose an abstraction shift towards ecosystem curation as a top-level driver for GI processes in order to balance the need for novel genetic material (breadth of search) with exploitation of high-utility regions (depth of search).