I doubt the designer of the font in that image was trying to make the letters look different sizes for fun. If that was the goal, though — great, mission accomplished!
You say it's wrong to judge overshoots by these examples because the text is too large. Okay. Show me an example of appropriately sized type that looks more _uniformly sized_ with overshoots than without, and I'll agree with you. (Remember, the justification for overshoots is uniform optical size, not playfulness!)
First off the bat, I dislike the FedEX typography at least as much as you; it is not a very sensibly, sensitively done design. With that out of the way, let me start with a few quotes.
The first emphasizes that overshoot is just one of many aspects of typography that require the designer to tweak forms; the platonic, geometric ideal has to be adapted to the realities of human perception:
Typefaces are born from the struggle between rules and results. Squeezing a square about 1% helps it look more like a square; to appear the same height as a square, a circle must be measurably taller. The two strokes in an X aren't the same thickness, nor are their parallel edges actually parallel; the vertical stems of a lowercase alphabet are thinner than those of its capitals; the ascender on a d isn't the same length as the descender on a p, and so on. For the rational mind, type design can be a maddening game of drawing things differently in order to make them appear the same.—Jonathan Hoefler & Tobias Frere-Jones [1]
This fact has been well known the world over and throughout the ages; for example, in the game of go, the stones, which are black and white, are of slightly different sizes (black slightly larger), to give the appearance of being the same size [1]
Likewise the Parthenon in Athens: "[The Greeks] achieved global perfection through deliberate departure from local precision. Minor geometric irregularities were incorporated by the architects to enhance the beauty of the building. It is paradoxical that these modifications create the impression of great geometric perfection, even though they involve deliberate departures from strict regularity." [2]
As for type design, the maybe most geometric-in-appearance modern typeface in wide use, Futura by Paul Renner (1927) is not perfectly geometric: The design of Futura [...] makes subtle departures from pure geometric designs that allow the letterforms to seem balanced. This is visible in the apparently almost perfectly round stroke of the o, which is nonetheless slightly ovoid, and in how the circular strokes of letters like b gently thin as they merge with the verticals. Renner's biographer Christopher Burke has noted the important role of the Bauer Foundry's manufacturing team in adapting the design for different sizes of text, a feature not seen in digital releases.
As becomes evident from the last sentence and as is widely known by graphic designers, many / most digital typefaces forego adaptation to different perceived sizes (I'd say that is one major culprit in what makes the FedEX logo so awkward: they just took a design that would have worked well for a magazine headline and blew it up to cover a room-sized area). In so far overshoots are not a holdover from metal type; rather, digital type through it's negligence of established designer sensibilities has produced awkward and ugly results that more traditional craftspeople would've been ashamed of (this BTW is quite parallel to the discussion around putting two spaces after the full stop).
I refuse to show you another image to which you will doubtlessly just reply "meh, overshoot bad". It's also a matter of taste.
You say it's wrong to judge overshoots by these examples because the text is too large. Okay. Show me an example of appropriately sized type that looks more _uniformly sized_ with overshoots than without, and I'll agree with you. (Remember, the justification for overshoots is uniform optical size, not playfulness!)