... There's a perception that lenses designed for the smaller format will always be lighter. ...
It seems that most of the EF-S lenses which appear to be lighter than the nearest full frame equivalent are wide angle zooms with a different design to FF lenses; a design with a short back focus where the rear element is closer to the sensor. With telephoto lenses this design advantage apparently doesn't exist.
I thought thaqt issue had been resolved years ago: the size advantage for lenses of equal focal length and aperture when designed for a small format is essentially when the larger format lens needs to be a wide angle design while the smaller format uses a normal or telephoto design, or when both are wide but the larger format lens is even wider. In the telephoto regime, the only advantages I know of to designing specifically for a smaller format are
(1) aberration corrections can be optimized over the smaller image circle, potentially getting better IQ there at the expense of worse in the unused regions outside the smaller format's image circle but inside the larger format's image circle.
(2) lens hoods and anti-flare baffling that is tighter, blocking light coming from outside the smaller image circle but within the larger image circle.
Both could be true for the Olympus 150/2, 90-250/2.8 and 300/2.8, and for the recent Pentax DA telephoto lenses.
It would be interesting to see a comparison between the Olympus E-620 with the Zuiko 300/2.8, and the Canon 50D with Canon 300/2.8 IS. The Olympus sensor has the edge in terms of resolution, but the Canon might have the edge in terms of noise.
Of course per pixel noise comparison are of little direct practical relevance when pixel counts are different: at least the higher pixel count image is entitled to a greater degree of NR processing so that resolution after NR is equalized. (The default high ISO NR of some cameras seem effectively to do this, holding the line on noise levels so that res. declines as ISO increases. That makes sense to me so long as it is only a JPEG default, not mandatory.)
However, if we crop the E-620's sensor to the 3:2 aspect ratio of the 50D, that edge in pixel density becomes so small as to be irrelevant.
It has no effect whatsoever if the objective is equal pixel count on equal field of view. A 3:2 image from a sensor with 4.3 micron cell spacing (12MP 4/3 sensors) is still smaller than a 3:2 image from a sensor with 4.7 micron pixel spacing (50D, 500D) by a linear factor of 4.3/4.7, about 0.9, and so one gets the same framing with a focal length 10% shorter. Not huge but my main comparison was the far more clear-cut gap between 4/3 and 35mm, not the modest 10-20% linear size gap between 4/3 and EF-S.
The question of different aspect ratios is partly why I threw in the Panasonic GH1 with its wider "multi-aspect ratio" sensor that optimizes image circle usage for various shapes. 3:2 is given by cutting a wider but lower rectangle out of the same 21.5mm image circle, and 16:9 even wider and lower, all about 12MP. The options are 4000x3000 for 4:3, 4128 x 2752 for 3:2, and 4352x2448 for 16:9, so the sensor active area is 4352x3000 pixels, 13x18.9mm.
I hope for this multi-aspect ratio option in many Micro Four Thirds cameras; it will kill the debate over which aspect ratio is most useful.