The more consistent the game, the more important the skill (which is good). Chess is an example: if I play 1000 games against, say, Anand, I will lose every single one. But Chess also gives the example why consistency is not good: if you look into, say Ruy Lopez Spassky Variation, it sets the best moves up to the 18th move (!), and if you want to play at high levels, you must have it all well memorized.
And chess is a much more complicated game than Gwent! Gwent is in fact quite simple (computationally): you have at most 10 cards in hand (plus pass), of 25 possible in the deck, and you make usually around 14 moves in a game. (I am actually surprised that no one has tried to analyze Gwent in a computer before; I assume they would do very well.) If the consistency of the game is high, people will soon be able to figure out that when playing Bear, the optimal outcome is always to start with Bearmaster -> Maiden -> Restore (if against deck X) or Raging bear (if against deck Y). And so the gameplay becomes stale.
Also, when games are more consistent, having better decks means you will win a huge percentage of the time. In the extreme case, if you only have 10 cards in deck and you draw them all, a better deck should win all the time. Which makes the meta stale. (To expand further: one thing is to play a tier 2 deck that I know well and opponents know little, if it has an average 45% win-rate. Another thing completely is if the average win-rate is 30%.)
I am not speaking in favor of Create here, there are obviously other ways of not making the game too consistent. (Nova and Shupe are another way they tried this.) But Gwent in design is already a very consistent game, since you often draw almost the entire deck each game, starting with 10/25 + mulligans, and drawing 3 more (plus mulligans). And CDPR introduced early on a lot of tutors and thinning and such. (Which I wonder if it was not a mistake, but it is already done.) (For comparison, in MTG you draw 7 cards of a 60 card deck.)
Again, I'm not arguing for RNG, I'm just saying that people that complain about RNG (be it create RNG, card draw RNG, etc.) and complain that the game is boring/stale should know that one often works against the other.
In fact, now we look back at pre-midwinter as a golden age of Gwent, but I remember when Midwinter was announced, a lot of people (streamers, etc.) were saying that they needed to do this, otherwise the game was dying. And if it is true that the player-base was shrinking, I wonder if it is not because although the game was very skill-based, each game didn't feel a lot like the next.