The author didn’t consider that a changed behavior - the inlined image already had a dependency on http to be downloaded in the first place. No, this is all on the original author.
I argue that the inline image had no such dependency, as it was an internal component of the control and could just as easily be used in a non-http context. By loading the image through a http request, the image component is effectively being injected with the http service in order to operate.
The correct refactoring is one the author fuelled with coffee, which is naturally the real lesson here.
