Stockfish is wrong in this analysis

@YALturKManI
The first position Stockfish evaluates as +0.2 from my side and not +12.
Though the second position is illegal, but White is clearly winning. For that play Stockfish's best moves and you will find Black in zugzwang position after 30 moves and Black would be forced to capture the White Queen on 8th rank otherwise White will win since Black has less space and after only 8 moves from the position, with best play, Black loses it's Queen.

@Akbar2thegreat when I played the best moves, I got a queen-queen trade on move 23

@YALturKManI
Probably because of different depth in our analysis.
Still, go ahead move 23 and if you are free for another half an hour it would be enough to arrive at a result or it would be clear from the position that White is winning.

lichess.org/analysis/QQQQQQQQ/PPPPPPPP/4P3/5k2/8/7K/8/8_w_-_-_0_1
I eventually got into this position on move 85.
@Akbar2thegreat let us not turn this into an argument
edit: also for the first one I still got this:

+12.0
Stockfish 10+WASM
Depth 71/99
now this:

+12.0
Stockfish 10+WASM
Depth 79/99, 431 knodes/s

@YALturKManI
Actually Stockfish doesn't understand fortresses and so it gives wrong evaluation.
For example take the case of the mentioned position which you shared.
The aim of Black king is to remain on the centre squares of Black side not touching the edge squares, that is, d7, e7, d6 and e6 are the central squares. Because if the king goes away to one corner then White will sacrifice it's Queen with Qxg4 (if Black king goes to the right of c-file from it's side) or Qxa4 (if Black king goes to the left to f-file from it's side).
So, with this principle if the Black king remains on d7, e7, d6 and e6 squares then White cannot force a victory and if White sacrifices the Queen then the Black king is not far from preventing Queening of the White pawn.
Hence, it's a draw.
Stockfish's CLOUD analysis shows it as a draw while beyond CLOUD it shows +12.

For the second one, have you saved the moves. If yes, then make it's study and request it's computer analysis there and share the link. If not, then I have done where without any bad move by Black, it results in win for White.
The problem is we cannot challenge Stockfish here from the position since the position is illegal.
The study that I created: lichess.org/study/Hu6G8Gdy

And so the spew of random numbers continues...

#8
"Since White's queen can't possibly cut through that logjam of pawns, the only way he could ever win is to play QxP...one of them. Furthermore, it presumably has to be the a4-pawn, since otherwise Black would have two recaptures available (thus making his coverage all the easier). What this means is that Black just has to keep his king within reach of a potential runaway b-pawn and no win will ever be possible."

I think the mildly interesting point of the OP is that humans use a type of logical/abstract reasoning to solve this which chess engines just don't master (yet) - such as the series of observations described above. Engines are better at chess than humans, but here still are problems where humans get the right answer within minutes, whereas engines calculate for hours and throw up pointless evaluations - rather than just proclaiming it a draw.

@Panagrellus
Well, there are many puzzles that Stockfish fails to solve but humans have solved!
You can find them by searching on YouTube 'Difficult for Humans but Impossible for Computers chess puzzles' and there is a play list of Chess with Suren channel that contains all such.

Rather than staring at the engine numbers use your brain!!