There is no difference between the two types of encounters.
Bottom line: they are both encounters. With the same creatures and scenario, which can happen, they both have the exact same amount of risk.
The rules do not change, just because the DM has a different path to an encounter or that he is creating the encounter on the fly instead of ahead of time
<snip>
A surprise round is just as risky with the same PCs with the same capabilities and the same NPCs.
There are differences. The differences lie in the who, what, where, and when of the encounter and how many of these factors the PCs can control or influence.
<snip>
The resolution mechanics (rules) do not change. The circumstances of the engagement CAN change dramatically, providing a telling advantage or disadvantage to one side or the other.
In order for the players to be successful in scouting, etc., it requires the cooperation of the DM. If he makes the DCs too difficult, the group might be back at the 30% success level. If he makes it too easy, then the PCs are getting easy XP under the illusion that the PC actions actually mattered. But in reality, the DM was helping.
This depends on what the resolution mechanics are for the PCs' prep. If, for instance, the mechanics allow them to create reliable hidey-holes from wish they launch an ambush, then successful ambushing is not just a gift from the GM. It is the players, via their PCs' declared actions, building up and then deploying resources to aid them in encounter resolution.
I think the higher the level of magic in the game - which means the more wacky the basic cosmology and menagerie, and the higher level the PCs and their opponents - the harder this becomes to fairly adjudicate. D&D is more wacky in this respect than (say) Runequest or even Rolemaster, but especially at lower levels I think the obstacles to fair adjudication aren't insuperable. For instance, if the party outfit themselves with darkvision capabilities, and then launch a night-time raid against a group of camping 0-level humans (eg a squad of soldiers), the GM can fairly adjudicate that the humans can't see in the dark. In AD&D, if the soldiers have a MU companion, the GM is expected to make random rolls on the spell tables to see whether or not that NPC has an infravision spell available for casting.
I'm guessing that ExploderWizard runs games mostly at low-to-mid levels (but not above 10th level). Maybe I'm wrong in that guess, but in my experience it's at those higher levels where GM decisions about what magical resources the NPCs have available make a huge difference to the prospects of the players' success, and become hard to adjudicate in an impartial fashion.
The DM does not cooperate or punish players when he or she sets DCs (or encounters, or whatever else). The DM does these things (or should) based on what makes sense, regardless of what the PCs do. I.e., if a particular check is hard, it's going to have a hard DC value. The particular level of the PCs, or their prof bonuses, or whether or not that check/encounter is hard for them isn't relevant. The challenges are there to exist in the game world and adhere to the rules of the game.
I don't really agree with this.
First, whether DCs should be set "objectively" or on a relative basis is a matter that varies from RPG to RPG and from table to table. 4e tends to favour "relative" DCs, for instance, and that is not a particular problem. It just doesn't sit particularly well with sand-box style pre-prep.
Second, the GM making decisions based on what "makes sense" often isn't a very good guide. Does it make sense that the wizard friend of the soldiers would have learned a darkvision spell to deal with night-time ambushes? Or that s/he has got unlucky and never managed to get that spell in his/her spellbook? This is why Gygax's AD&D emphasises random determination of spellbook contents, but that has its own oddities (how did someone who knows only Affect Normal Fires, Read Magic and Pyrotechnics get a job as a caravan guard?). And in any event 5e doesn't seem set up to support random selection of spell (eg there are no numbers next to the spell names, unlike the original AD&D spell lists).
There is more to impartial adjudication in sandbox style RPGing than simply "doing things based on what makes sense".