Display one of two questions randomly in a loop & merge, but not always the same question

Hello,
Is there a way to do a loop & merge in combination with randomization, where the loop resets the randomization each time?
A short elaboration of the problem:
Survey takers are taking a 50/50 (win/lose) gamble or not taking a gamble. The 50/50 gamble changes each loop. If they choose to gamble, they see whether they won or lost, this is a text/graphic question in my setup.
There are thus two different text/graphic questions that they can get, and I want them to get these randomly. However, with loop & merge, the randomization occurs only once; i.e., they only see the win question OR only see the lose question after each gamble in that block. Any work arounds?
A long elaboration of the problem, more detail on how I have been dealing with it:
Survey takers are taking a 50/50 (win/lose gamble) or taking a guaranteed option.
e.g.: Where, Field 3 is a set of guaranteed amounts, Field 2 is a set of amounts to lose, and Field 1 is a set of amounts to gain. All sets consist of 90 numbers.
QUESTION 1:
Choose one of the following. The gamble decision has a 50% outcome for either of the presented outcomes.
⏺️ Guarantee: ${lm://Field/3} ⏺️ Gamble: Gain $${lm://Field/1} or Lose $${lm://Field/2}
QUESTION 2:
You lost $${lm://Field/2}!
QUESTION 3:
You gained $${lm://Field/1}!
QUESTION 4:
You gained $${lm://Field/3}!
-------------------------------------------------------------------------------------------------------------
If they choose Guarantee, they should be skipped to question 4. If they choose gamble, they should be see either question 2 or 3 randomly, but not both.
I have tried several things thus far but have not found a solution.
Where I am now:
Skip Logic
Question 1: Skip to Question 4 if Guarantee is chosen
Question 2: Skip to End of Block if Question 2 is displayed
Question 3: Skip to End of Block if Question 3 is displayed
Question 4: Skip destination
Randomization/Order
1) Question 1
2) Random Subset: [Question 2, Question 3]; only present 1 (have also tried 2); evenly present
3) Question 4
-------------------------------------------------------------------------------------------------------------
I am also open to completely changing how I have set this up if I can get the same effect. The main thing is that they get the result of their gamble/guarantee after each question.
I guess the long way to do this would be to create a question for every single decision. Down the line, I may want to count up what participants earned as a result of their 90 x 3 (there are actually 3 within subject conditions) gambles--can't imagine this is easy with the loop and merge setup I've got.