Swift Data usage with global container and ephemeral contexts

Using Swift
1

I’m currently using Swift Data with the following pattern:

  1. ModelContainer is initialized once in the global scope as a global variable.
  2. The container is injected into my app using .modelContainer(container) and accessed in views via the env variable.
  3. For all other use cases, including tasks and various system callbacks (async or not), I create a ModelContext on the fly from the global container and discard it after the specific execution path is finished.
  4. ModelContext instances and their associated models are not shared between tasks or jobs, but they may be used in async tasks without being tied to a specific actor.

Is there anything inherently wrong with this approach?

I’m unfortunately seeing fringe errors, such as EXC_BAD_ACCESS, KERN_INVALID_ADDRESS, and EXC_BREAKPOINT, mostly related to Swift Data. I’m trying to sanity-check my basic approach to determine whether the issue lies in the overall design or elsewhere.