How big is the difference between optimize = "size" and optimize = "speed"?
optimize=size generally produces binaries about 15% smaller than optimize=speed, but the exact figure depends on the target platform.
There are no Godot-specific benchmarks I know of for different SCons flags, but the general figure with GCC/Clang is that -Os (the default for optimize=size) is 20% slower than -O2 (the default for optimize=speed). However, if your performance bottlenecks are not related to the CPU (such as GPU or I/O), then the difference will be lower.
