I think for him it's more about being able to "own" dwarf fortress. If he open sourced it, large teams of people might be able to develop it much faster than he could, and if he doesn't really want to manage an open source project, that might end up in him losing control/maintaining a less popular game than whatever gets forked out. It's also widely suspected that the codebase is really bad, and he might not want to manage a clean up of it that makes it 'better' but which results in him not understanding it as well.
If LCS is any indication, the code in DF will be a horrible mess.
That being said, LCS came out in 2004, so presumably Tarn has improved over the years.
In practice that likely means that the code is not all that well organized and planned out. Doubly so if he and his brother have been the only ones who have touched it -- there's a lack of outside influence that would drive one to clean up their code. I of course have not seen any of the source code for DF, so I this is purely my own hypothesis.
At one point I tried to edit Nethack to make it impossible to die. It was kind of shocked by all the random checks strewn throughout the code.
[1] https://github.com/DFHack/df-structures [2] https://github.com/DFHack/dfhack
The saddest part of Tarn's decision to put the game on Steam is to make enough money to pay off medical bills for his brother.