The last factor which contributes to the lethality is the aspect of Vaughan's design which I enjoyed the most: the foes are largely unknown and unknowable. Unlike many adventure products, Vaughan is constantly reaching for new monsters, new templates, new classes and applying class levels to monsters in a manner many GMs and players will be unfamiliar with. This means that unlike most adventure products, the players (as distinct from the PCs) usually never know what they are up against and never know what the foe's capabilities actually are. This aspect of the design adds immeasurably to the difficulty of the Slumbering Tsar Saga. If "knowing is half the battle", then the PCs start each battle in Slumbering Tsar on the losing side.
In terms of lethality I think this can be balanced based upon your judgements, but each part is more or less difficult to do. First, if the overall challenge level is out of bounds, just leave it until the PCs are a level or two higher. That might mean 7th-12th level instead of the suggested 6th-10th.
Random encounters in a territory are probably balanced with the area. However, while I don't know how big the populations are I like to integrate random encounters just in case 3 goblins ones in a row doesn't also end up in a full tribe still left in the lair. Shrinking the population matters even with random scouting patrols.
The quote above is about increased lethality due to not knowing and I think that's best mitigated by adding a information / rumor system. Creatures in shared and neighboring territories, especially ones in common alliance, know about each other. Finding out who they are, what they can do, what treasure they have, what they know (second hand, of course), and anything else in their history can be part of the game. It can also make combats considerably easier. Of course, creatures can more or less lie too and lead others into more dangerous situations, so this portion of the game isn't' necessarily sweetness and light either.
Besides the potential for more rumors to mitigate going into harder territories, the most difficult part of changing a finished module is in how the dungeon level by challenge level is actually laid out. The example ridge means high ground, so that may be the better territory to hold - treasure-wise too. But if the divisions don't hold in their own right, nothing is keeping the higher level monsters from swarming into more valuable, less defended territory, then the environment is going to change pretty quick. Part of this has to do with adventure design and may just be ignored altogether in this case. The key part to insure would still be distinguishable borders even if they're the proverbial "wrong side of the tracks" rail line. If you want to go further, you can redraw stuff, restat stuff, or even rewrite things, but there's a point where it stops being pragmatic.
From the review this really does sound like a great product. I count myself lucky to own the Rappan Athuk boxed set since it came out and it is one of the best big dungeons around. Unfortunately, I'll probably be picking this one up second hand sometime in the future as my gaming funds are currently strapped. Definitely on the To Buy list though. Thanks for the review.