If it's something I've prepped, and they're getting to it well after I prepped it, then I'm changing things mostly because the PCs have different needs than they did when I originally prepped [place]. If this is a place they're returning to (not included in your original question, but still feels vaguely applicable) there are helpings of verisimilitude and not-stagnation, too.
EDIT: And I prepped it the first time because I thought the chances the PCs would go there made it worth doing.
I think this is really due to the fundamental difference in how these games do PCs and mechanics. In D&D, PC capabilities vary wildly as they progress -- the easiest to demonstrate is spellcasters, who gain massive changes in what's possible and not just in power. This is true of other classes as well, it's just most obvious in the spell lists. A character can switch from having to walk across a continent to doing so instantaneously in the blink of an eye at the table -- the player just levels up the PC to get teleport. This radical shift in capability (again, present in all classes), means that unused prep needs to be updated, if only to account for new abilities. Elsewise, prep must be discarded as not something that's relevant to these PCs anymore (a chasm, for instance, goes from a dangerous obstacle to a triviality very suddenly). In Blades, though, this doesn't happen -- PCs do not have sudden shifts in capability, they just improve at doing what they already could do.
The upshot of this is that, in D&D, prep is absolutely necessary, if only to account for the difference in capability. This applies to sandbox prep as well, as different areas/locations are prepped with different capabilities in mind (this is a challenging area with powerful opponents, this is a less challenging area with weaker opponents). With Blades, though, this isn't necessary -- threats are more about what's happening rather than quality of the opposition/accounting for increased abilities. PCs get more competent, but don't add entire new categories of capability.
I think this difference -- how you account for PC capabilities -- is a critical difference in how these games can and are prepped. Blades is fine with a thumbnail and an opening situation, because that's all that's needed to engage with play. When the PCs start interacting with the Billhooks, for instance, then what the Billhooks are doing will necessarily be an impediment to the PC's plans (elsewise, the GM will not invoke them in framing or consequence). They don't need to adapt, in any way, to the PCs, nor do they need to advance on their own because whatever they were doing is of little consequence to play, it's only what they're doing when the PCs encounter them. Now, that said, I'm not at all adverse to making a Fortune roll and advancing their clock as I introduce them, especially if that puts more pressure on what the PCs are trying to do (and this is a consequence framing). In D&D, prep absolutely has to account for the PC abilities -- either in adjusting previous prep to remain useful (which seems your approach), or by tiering threats in a sandbox so the PCs have a variety of challenges for whatever their current abilities are.