Miner starts digging, while rest of the seige clusters around the tunnel entrance, defending it. Miner randomly goes up or down a Z-level every so often to keep you guessing which Z-level to put your army, in case you find him. As soon as the miner hits Fortress, he begins to widen the entrance as fast as he can before he's forced to retreat. Seige only goes down the hole once the miner hits fortress. This solves the concerns you addressed.
The whole point of a "siege tunnel" or whatever is that the attackers go down in it. So when such a tunnel is dug, one of three things happen:
1) All the attackers go with the sapper. In this case, the scenario I described happens: bloodbath in the tunnel.
2) All the attackers stay at the entrance*. In this case, the tunnel does literally nothing for them. Thus a whole ton of time and effort will have been spent doing absolutely nothing, when it could have been spent doing something productive. Like, say...anything.
3) Some of the attackers go with the sapper, some of the attackers stay at the entrance. In which case both scenarios happen at once: one set of up/down stairs lets defenders be right where the breakthrough is no matter what z-level it's on, resulting in a bloodbath at that end, with the rest of the troops being irrelevant to the discussion.
Then the miner makes the tunnles 2, 3, n squares wide, allowing more goblins to rush in at once. If the goblins stay on the outside until the miner hits fortress and widens the hole, it lessens the time they actually spend during the tunnel. Worst case scenario the sapper gets intercepted by your own tunnel and get killed, which case all the goblins lose is a sapper.
A set of stairs close to the walls (or rather, all the walls since we don't know where the goblins will dig in) can help in rapid deployment, but would absolutely stink if something went wrong and the goblins made it past the breach, since they can now use them to rapidly spread throughout the fortress, turning it into an urban warfare situation.
*(by stay at the entrance, I mean don't go down the tunnel, one way or another. Hang out aboveground by the entrance, hang out a couple tiles inside, go attack the old-fashioned way, whatever.)
Let me try to put my feelings another way: basically, the big problem I have with this sort of siege-tunnel is that it's a solution without necessarily diagnosing the problems. As I see it, the problems with sieges are this:
1) Walls and/or moats make a siege irrelevant. Lock the doors, plant some plump helmets, and wait for them to get bored and leave.
2) Traps are overpowered.
3) Fortifications make Marksdwarves overpowered.
4) Champions will utterly massacre ten times their number without a scratch.
The way to fix the current problems with sieges is to reduce or nullify each of these problems. There are lots of solutions to these problems, many of which have been suggested time and time again, and all these solutions are a lot simpler than trying to create an AI that can actually dig sapper tunnels worth a damn:
1) DG's idea of making sieges cause unhappy thoughts. Wall-breaking/crossing + moat-bridging.
2) Nerf traps in some way (for which there are as many different suggestions as there are posters).
3) Siege engines to knock away fortifications. Bolt-resistant attackers (such as siege towers). Outright nerfs to ranged weapons.
4) Make champions harder to get. Make goblins not suck (seriously, unarmored, non-elite wrestlers versus ☼steel☼-clad axedwarves?).
Now tell me that it would be simpler to write an entirely new AI routine that not only works, but that can deal with terrain that even experienced players have trouble with than doing these things would be. And tell me how making sappers instead of these sorts of things would help when digging is impossible, such as when there's an aquifer.
It would be simpler to write an entirely new AI routine that not only works, but that can deal with terrain that even
experienced players have trouble with than doing these things.
. . . What? You asked!
Seriously though, if you read my posts you'll realize that the reason I brought this discussion up (notice it's not in the suggestions forum) is to bring attention to the fact that this tactic could be much more effective in the DF-verse than in real life, can potentially be done by both players and AI, and can drastically alter how warfare is handled. It was not intended as The Solution to the Seige Problem, though it does do help alleviate the problem. It's just another potential idea on how to enrich Dwarf Fortress by allowing as much freedom as possible when performing actions. If Dwarf Fortress was programmed with the "simplest way to solve a problem" mentality, it wouldn't BE Dwarf Fortress. We're talking about a game that allows you to simulate thousands of years of history here!