Walling off being something Toady doesn't intend to be possible leads me to not take it into account when it comes to suggestions not about fixing it.
I know that you're not supposed to be able to wall off the surface, especially from sieges, but I'm not sure that extends to things like FBs. Taverns actually create a pretty decent reason to keep your fortress open. Some creatures are already functionally invincible against anything but obsidian/ice casting or cave-ins. It'd be game-breaking to have steel blobs or fire-generating creatures that can pass through walls. (At least without having pathing exploits that make sure they tunnel into things designed to cause cave-ins, but even that would create more annoyance than solution.)
Risk also not need come with reward to make it worthy of inclusion as challenge is it's own reward (the deepest depths)
But the deepest depths do come with a reward. It's what you mine on the way down. Your risk is the HFS, which was always pretty much supposed to be the Game Over screen. It's just been changed from absolute you lose to being a blatantly unfair fight that can only be won through severe exploits.
The fact that there is nothing for you down in the depths but endless functional FB++ monsters was always a "OK, nothing to see here" sign. People do it merely because it's so hard that it becomes an epic tale by default.
it would be reasonable to expect a tall tower to be visible from range (ties to this suggestion). A tower should be fairly easy to detect and give the fort a reputation.
This is more interesting, though.
Toady's quote indicates the general problem, but if we assume that towers are always constructions, you could make some forms of reasonable machine extrapolation of player creations by simply counting the number of constructions above the "natural" z-level plane of a fort.
That is, if you have 25 tiles of construction at an elevation 5zs above the designated surface level, as indicated by starting soil layers, then another 25 constructed tiles at an elevation 6zs up, 25 constructed tiles 7zs up, and so on with maybe 29 constructed tiles at 14zs up (balcony), then it would be reasonable to label it a "tower", regardless of how well the game can properly understand what it is the player is building or its intent. A "pyramid" might be the same thing, but with 1 constructed tile at the top, 9 one z below, 25 one z below that, and 49 the next z down, etc. (Provided it was within a certain horizontal distance, as well to make two different towers seen as two different towers... again, kind of tricky, but doable if you count something like support to see if any of the constructed tiles support one another without having to go through natural tiles, first. Although whether it is counted as two towers in a single embark or not may be irrelevant for game purposes.)
By extension, open-air pits should be notable, as well, as measured by whether or not stone layers are considered "subterranean". If the player excavates a whole chunk of embark down to the first cavern, that's notable as being a "pit".
Giving players the option to arbitrarily declare a given random large contruction something other than a tower makes sense, as per Toady's comment, but the game would just be checking for the "unnaturalness" of how much you have terraformed the map before it considers letting you call something a tower or castle or great statue or whatever.