Background: I am workshopping the dungeon from this question, and this answer suggests that I should have the rogue make a Disable Device check up front in order to figure out the trap's workings. However, the Disable Device DC on these traps is variable, depending on how the rogue elects to go about disarming the trap! Is it OK to have 2 Disable Device rolls for a single trap at possibly different DCs (and different locations, for that matter), one to determine the workings of the trap and a second to actually disable it?
It's okay if the players know what's happening
If the DM says, "Make a Disable Device skill check," and the PCs expend resources to make that Disable Device skill check, they'll expect, afterward, a disabled device (or, y'know not, depending on the result). Unless warned beforehand, the PCs won't expect that they'll need a second check (or more) to actually disable the device. That's mean and liable to create hard feelings.
Be up front: "This is a trap like you've read about in rogues' school where the first part is so incredibly complex, even understanding how to disable the device requires a Disable Device skill check," the DM says. "Once you understand this trap, then you can really disable it."
A pair of alternatives
First, instead of making the Disable Device skill the only relevant skill, the DM could have PCs make Knowledge (architecture and engineering) or Craft (trapmaking) skill checks to determine the working of a multistage trap. Success may grant a bonus to the future Disable Device skill check or lead to the party safely to the trap's next stage, while failure means suffering a penalty or triggering a minor trap, a tampering countermeasure.
Second, consider the optional rules for complex skill checks, looking carefully at the specific rules for complex Disable Device skill checks. While I've not used these rules in Dungeons and Dragons 3.5, I have used slightly different iterations of such rules in both first and second edition Spycraft campaigns where the rules work quite well.