The (an) engine creates various 'textures' that cannot be 'eliminated'. To some, you may have access through various channels, to others not. So, no idea about OpenGL ES but if it was OpenGL there'd be a default framebuffer created by the context, and possibly a number of others used by engine for rendering of various effects. A framebuffer can have various textures attached, colour- and depth- are usually always there, others possible. The OpenGL framebuffer has at least two pairs of such textures in OpenGL, a front- and a back-buffer (colloquially called frames sometimes). While one is being displayed, the other is being drawn to. In Vulkan the application can choose how many such frames there are in-flight, often times there are three of them in a ring buffer, so to support different display modes.
Apart from that, there can be other textures, for rendering shadows (multiple levels possible), for a skybox (6 of them), the blue Godot wif hif teef fticking out, fonts and symbols in various sizes, and whatnot the engine might have in mind.
tl,dr: It is not that you don't have nothing even in a blank project, as Megalomaniak said, at least there must be the possibility to show you the dark screen with nothing in it, and that needs textures and memory on the video card. Though I doubt the default framebuffer is being on display there, at least not if it is OpenGL.
Edit: ninj'd, there you have the answer from the cracks, sky material, textures for font rendering, symbols. And 1.8mb is really not a big thing.