I have been thinking a bit about this, and I think this is might have some important challenges:

• being able to inject arbitrary patches is equivalent to modifying conandata.yml on the fly. The same principle could be used to modify also the conandata.yml "sources" origin.
• This injection, applied to any recipe existing in Conan cache or Conan servers, to build some dependencies injecting patches, means that the recipe-revision is not changed at all, it cannot be changed, as that implies a whole new export
• If the recipe-revision is not change at all, then there is no traceability, reproducibility or correct package_id computation, changing the applied patches will not trigger a new build, and even if the build was forced it will never know the corresponding source patches that were applied

So the only possibility that I think would make some sense is to have an extra conandata.yml location for conan export/create time from conan-center-index fork, similar to the settings_user.yml approach, when it is possible to load extra patches from another location. But I think this sounds a worse alternative than just adding the patches in the original conandata.yml and have a clear diff. Was this the original idea of this ticket?