As Pumbaa pointed out, many strategies struggle or fail with the case where the 2 false coins have the same averageopposite weight as the others-offsets that can cancel each other out.
You divide the 24 coins in 8 stacks of 3 coins. (make sure to mark the coins so that you always know what stack they came from). Then we compare the stacks in 4 pairs. (stack 1 versus stack 2, stack 3 versus stack 4, etc...). Out of these 4 measurements, 0, 1 or 2 can return an inequality (where the scale is not balanced). Let's handle these 3 cases:
Case A: Two of the stack-comparisons were unequal: That means that we can dismissconfirm the other 12 coins as genuine and look for one false coin from each unequal measurement. To find a false coin in such a stack-pair, we remove one coin from both stacks, and swap one coin to the other side. The resulting 2x2 coins are compared. This measurement will reveal that the false coin is either among the 2 removed coins, or among the 2 swapped coins or among the 2 other coins. One more measurement (with the help of a genuine coin) will reveal the false coin. If we do that for both stack-pairs, we get a full solution in 8 measurements.
Case B: One measurement is unequal: This means that both false coins must be among the 6 coins of that measurement. ThatFinding them takes 5 measurements: Comparing coin 1 and 2, then coin 2 and 3 etc... (I'll leave it to the reader to proof thiswork out the details. Consider that atHint: At least 1 measurement will return equal, and that every unequal measurements includesmust include at least one false coin). That's a total of 9 measurements
Case C: All measurements return equal, that is only possible in a few special cases where both false coins are part of the same measurement (but we don't know which one). For the next5th measurement, put stacks 1 and 3 on one side and stacks 5 and 7 on the other (so 6 coins on each side).
Case C1: If this 5th measurement returns equal, then theboth false coins must be in the same stack, canceling each other out with opposite weight offsets. Let's find them in 4 measurementsOur 6th measurement: From stacks 1 to 56 put one coin on each side of the scale (so 56 coins on each side):
Case C1a: If the scale returns equal, then the false coins must be either in stack 6, 7 or in stack 8. Put two coinsCompare a coin of stack 6 on one side and two coinseach. And again another coin of stack 7 on the othereach. If this returns equalBy now, thenthere's only 2 possible answers, comparing the false coins must be inright coin from stack 7 or 8. Next we compare 2 coins (A&B) of that stack with the third (C) and a genuine coin. If it measures equal, then A&B are false. If unequal, then C is false and one more measurement of A vs. a genuine coin will prove either A or B to be false. That's should eliminate the last doubt: 9 measurements in total.
Case C1b: If the scalemeasurement from C1 returns unequal, then the false coins are both in one of the stacks 1 to 56. If you move the same 1012 coins to one side of the scale and compare with 10 genuine coinsthe other 12 (you have 14 confirmed so farcertainly genuine) coins, then you can dismissidentify one coin of each of the 56 stacks as genuine. Now you just need to identify the right stack, which is relatively trivial in 2 measurements (especially since you know which coin one will be overweight and which one underweight):
First let's reduce from 6 stacks to 3 stacks: Compare the candidate overweight coins from stacks 1,2&3 with 3 genuine coins. For the final measurement, you put on one side the candidate overweight coin from one stack with the candidate underweight coin of another stack and compare with 2 genuine coins. Total: 9 measurements