Well, dfhack or whatever you call it sounds like a boring way to play, and savescumming seems terrible. I don't mind if my fortresses die; with each death I gain one extra ounce of knowledge!
Savescumming can be interesting. I don't use it to save myself from my mistakes, if I killed the fortress, it is dead. But I like to use it to study how things could go differently. In a memorable instance, I had a fortress that was going great, but with no defense. No militia, no bridges on the entrances, no levers to close those, etc. 30-40 goblins came and killed everyone. I had a backup save from 3 months earlier (I was playing in the school library, with saves emailed to myself, so i already had a backup), which I reloaded, and tried again. I immediately drafted everyone I could afford, and started intensive training. Goblins killed them all. Repeat 5x or so with variations in weapon selection and training. Same results. I tried everything, installing traps (not enough time to make enough traps), intensive training of various sorts. None of it was good enough. I proved to my satisfaction that there was no way to defeat the goblins in that time span, short of installing some doors and locking them till the goblins leave. But then I discovered that I was wrong. By digging an alternate trapped entrance corridor in soil layers with two long zigzag trap corridors ending in upwards hatches, I could manually path invaders back and forth across the same series of hatches, potentially sealing them in specific portions for trap maintenance (unnecessary in this case), using a small number of traps to kill an arbitrarily large number of invaders.
Once I had solved the puzzle I abandoned the fortress, and started over, with more defensive planning integrated from the beginning. Savescumming just gave me the opportunity to fully explore the challenge I was presented with.
Similarly, sometimes I'm curious about some aspect of game mechanics, usually related to minecart and/or fluidic contraptions. Those tend to be time consuming to build properly, which prohibits free form experimentation. For those cases, I have a backup save of an already established fort with dwarves modded for extreme speed, so almost not maintenance is required and designations get completed almost instantly (I know I could mod dwarves to not eat etc to make maintenance truly zero, but that seems like cheating too much. If you can't keep a fort fed and boozed with speedhacked dwarves, well, that's not something to be proud of). I like to think of it as being like minecraft's Creative mode, where you can build to your hearts content and try out ideas. I could do the same experimentation in vanilla dwarf fortress, but it would take many dwarf years and result in a mess of misdesignated passages and flawed designs. I prefer to implement complex designs in real forts
after I kind of figure out how the whole thing is going to work. Not that I make it a smooth and polished design before construction, there's always bits of abandoned cart track and such things in my forts, but I at least like to get a workable skeleton that I can continue to modify.