Now this I strongly disagree with. The idea that proceeduraly generated conted will be inconsistent (or more inconsistent than any other RPG campaign) is something I don't find true.
My experience has been different. However, it is not a question of which experience is more relevant. It's about understanding our respective circumstances and why we make the creative choices we do.
An important consideration is how long the referee plans to use the setting that is the focus of the sandbox campaign. If just for the one campaign, then inconsistencies are less likely to matter. This also means that ease of prep is a primary consideration, which makes the use heavy use of procedural generation a wise choice.
However, for someone like me who uses the same fantasy setting across multiple campaigns, consistency is crucial. So I use procedural generation differently compared to someone solely concerned with ease of prep.
Proper note keeping solves that pretty well, same as any other campaign.
I addressed this. It's not a problem if the referee uses procedural generation as a basis for preparation. Using notes to maintain consistency is a form of preparation.
When the referee or player (in the case of a solo RPG) is not involved and the result is accepted 'as is'; that is when inconsistency issues arise.
This product, released in the 1990s, is an example of this happening.
Even with the player or referee actively involved, such as with the use of notes, the use of procedural generation in a process where you only generate what you need when you need it has additional consequences. It is straightforward to keep new items consistent with what has already been established. Often, retcons are needed to keep older items consistent with the information that new elements have revealed.
How much of a problem is this? It depends on the nature of the setting and the complexity of the circumstances. But generate enough detail over time, and then retcons are inevitable. Players who play in successive campaigns in the same setting will notice, and some will care about the resulting inconsistency.
An example is my Majestic Wilderlands. It originally started out as Judges Guild, Wilderlands of High Fantasy. Unlike most published settings, it had a wealth of local detail but not much overview material like setting history or regional overviews. As a resul,t most referee using the Wilderlands came up with their own histories and their own regional overviews.
For me, around 1988, after a dozen campaigns using the Wilderland, my original maps were falling apart. So I decided to redraw the maps by hand using a homemade light table and large sheets of hex paper. In addition, I had a lot of additional details on culture, religion, and history, along with tweaks to the geography. I decided that in order to keep things consistent, I wouldn't redraw the original but a new version that reflects these changes. In essence, retconning the geography to better fit how I ran my Wilderlands campaign in the late 80s.
And it is no more limiting than having preset scenarios a la a DM curated setting.
Nothing is a free lunch. All creative choices have consequences, including the ones I use. These consequences are often not about whether there is player agency, but how player agency works out.
Ultimately, it's about matching the prep method to the campaign's goals and ensuring the players’ experience stays consistent with the world as it unfolds.