Big point swings are fine, as long as they are not achieved by one card with a huge variance. Dragon's Dream and Lacerate are high variance cards. These should be adjusted similar to Golden Froth. Gimpy Gerwin is another one with huge variance [...] Ridiculous design.
Not all high variance cards are inherently bad. There are basically three types:
1. Finishers (paper)
2. Tech (scissors)
3. Value (rock)
Finishers
Hubert, for example, is such a high variance card, but requires some setup to get the full value, unless you use Adda. Dragon's Dream falls into the same category. The card is somewhat risky because it requires Nivellen and a semi-long round with plenty units. A lot can go wrong in between. Value-wise, I don't think DD is an issue. It's just that you cannot always predict the card because it can fit in various decks. The high variance is not the problem here, but limiting it could be a solution, which still ignores the core issue.
Tech
Gimpy is different because it's a tech card that rarely gets insane values, unless you face NR or AQ. Similarly, Geralt: Yrden is a tech card that get get good value against Monsters, but can be struggling to find value against control decks. You could still make a case against Gimpy, but the card isn't problematic. One-trick-pony decks, like Draug and AQ should be susceptible to tech, otherwise they spiral out of control. Every deck has an Achilles' Heel. Draug can also be AoE locked by Auckes and AQ can be counter by herself in the mirror (ironically), but also wiped by Yen or Forktail. Both cards having a high variance, but also means to build your deck around.
Value
Value or value potential, in the case of high variance cards, are cards that need a condition to gain above average value. If the condition can be met easily enough, such cards could be included, like Lacerate. They are usually the most boring and value plays will always stick around in Midrange decks. They are usually not too problematic.
Closing Note
High variance cards is a bit like RNG, it's not inherent bad, but it does need to be implemented properly. Some cards are bad, some are not. Ironically, it's usually not the high variance itself that's the problem, but the condition therein, to unlock the cards full potential. In the end, we still need swing cards because that makes the game more exciting. Pulling off a Glustyworp should be rewarded and if Vran Warrior and She-Troll sticks around too, well, R.I.P. opponent. Wait, did (s)he just played Yrden? Damn.
This post turned out longer than intended.