this is a regression with color management, TIFF's were always being written as 8bit however the float buffer is assumed to be linear when converting from float to 16bit pixels per channel, so support for color management might need to be added here.