Thanks!

Do you see any reason for not using glBufferData() the same way for the vertex buffer? instead of glMapBufferRange() etc.

Much shorter :) not sure if the old way had any benefit

        glBindBuffer(GL_ARRAY_BUFFER, g_VboHandle);
        glBufferData(GL_ARRAY_BUFFER, (GLsizeiptr)cmd_list->VtxBuffer.size() * sizeof(ImDrawVert), (GLvoid*)&cmd_list->VtxBuffer.front(), GL_STREAM_DRAW);

        glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, g_ElementsHandle);
        glBufferData(GL_ELEMENT_ARRAY_BUFFER, (GLsizeiptr)cmd_list->IdxBuffer.size() * sizeof(ImDrawIdx), (GLvoid*)&cmd_list->IdxBuffer.front(), GL_STREAM_DRAW);

Didn't spot the possible simplification, but in this case it seems viable. glMapBufferRange can be more powerful, if used wisely. A possible future optimisation would be to double buffer the VBO and element buffer. This would allow to use the GL_MAP_UNSYNCHRONIZED_BIT and reduce the driver overhead and possible stalling.

We should aim to keep the example as simple as possible really.
If people really want to squeeze extra on the small rendering cost of imgui I'm sure this sort of people will have no issue with doing that.