We were molding a long, thin part with a U shpaed cross section - imagine a part that's 24" long, 1" wide, and maybe 5/8" deep (that is the direction of the open side of the 'U').
The part was attached to a panel only on the ends - there was no economical way to add a fastener to hold it down in the middle. No big deal if it warped concave (relative to the mating panel) when molded - the middle of the part would be in contact with the panel and the screws on the ends would suck the ends down and the part would lay flat. However, mold-flow analysis showed that the part would warp convex when molded no matter what we did with the gating and cooling. Uh-oh.
It is our policy that we design parts and it is up to the molder to give us those parts to match our prints. But when talking to the mold-maker he was going to cut the file right from out data and we knew that wouldn't work. So we violated our policy a bit and provided him a pre-bent part (using deform) that would then warp back to the part that we wanted in the first place.
Prediactably, the job got killed before we could see how it worked for real. However, mold-flow showed that the bent parts would warp back to flat, and (more useful) cast prototypes from an RTV tool of the bent parts warped back to flat just like we expected.
That was a few years ago. I have not had an occasion to use it since. Ed
BTW, I think Parel had an online tutorial which used deform or flex to change a sculpted handle. Though I do not personally agree with that (I figure model the part you want in the first place) I can't fault anyone for it if it works for them. Ed