Image 1 Caption: A cartoon the eternal inflation which is both a version of inflation and a version of the multiverse: a universe made up of some kind of pocket universes (see below).
Note eternal inflation and the multiverse are both highly speculative theories. The inflation paradigm in itself has a measure of robustness: it's resisted falsification since introduced in 1979 (see Wikipedia: Starobinsky inflation: History) though it has undergone considerable evolution.
Is it falsifiable?
Actually, it may fall into Flexibility: item 3. It has arguably passed one falsification test, but it's hard to think of another. Is one falsification test enough to be a scientific theory?
To explicate: the multiverse with varying ordinary physics has passed one non-trivial falsification test as discussed by, e.g., Martin Rees (1942--) in an article that yours truly has lost track of.
The explication of this falsification test is as follows the form of dialogue between Pro and Con:
This suggests they are roughly what they are because they were chosen somehow randomly and their only relationship is that they are consistent with life as we know it: i.e., they are ordained by the anthropic principle.
Now if they are were chosen randomly, there must be a multiverse to do the choosing.
As to the multiverse, it's not a scientific theory because it CANNOT be disproven.
UNDER RECONSTRUCTION below
The multiverse would be falsified if the parameters had special values.
In this case, we would try to find the logic of reality dictates these special values of parameters.
So the multiverse has passed a significant falsification test.
This does NOT prove the multiverse, but yours truly thinks that the multiverse does have to be considered seriously as a possibly true theory.
There is nothing recognizably special about the inverse (AKA reciprocal) 137.035 999 084(21).
But say counterfactually that the inverse was exactly 137 to very high accuracy/precision. That would suggest that the inverse was exactly 137. Would that falsify multiverse?
Well no. It could be that the fine-structure constant had to be an integer for some fundamental reason, but the integer could still have been chose at random. Note there seems nothing special about 137. It's a prime number, but not an especially interesting one like 2, 3, 5, or 7. The integer could just be the prime number 137 by chance.
On the other hand, if the fine-structure constant had seemed to be exactly 144 = 12**2, that might be considered a falsification of the multiverse since 144 = 12**2 seems special 144 = 12**2. Maybe reality is like the sixth grade where the multiplication table stops at 12 (see Wikipedia: Multiplication table: In modern time).
But maybe the proponents of multiverse could still find an argument against falsification if 144 = 12**2. After all, it could still just be an random integer.
Change some of the parameters of ordinary physics too much and complex structures such as life as we know it (including us) could NOT come into existence.
Actually, the anthropic principle is just a special case of proof by the fact that results imply causes.
In our present argument, we do NOT require that we exist, but only that the observable universe be sufficiently biophilic ("life loving") to allow complex structures such as life as we know it.
The life as we know it didn't have to be humankind.
A biophilic observable universe is like a lottery: there are a few lucky winners and many unlucky losers.
We're just lucky winners so far. A lot of random events went into making us what we are.
But there had to be a lottery for there to be winners. The biophilic observable universe is the lottery.
There is probably only a narrow range for parameters of ordinary physics to be biophilic.
So the multiverse is also a lottery in which probably there are only a few lucky biophilic pocket universes.
Most pocket universes are probably lifeless with only simple kinds of structure.
In fact, the parameters of ordinary physics do NOT seem to be super fine-tuned.
There are NO special relationships among them that indicate they had to be the way they are by some logic of reality that is is indifferent to life as we know it.
For example, say the fine-structure constant was exactly 1/137 instead of its actual value of 1/(137.035 999 139(31)) (see NIST: CODATA: All physical constants).
If the counterfactual fine-tuning held, some logic of reality dictates ordinary physics as we know it and life as we know it is an accident of reality---but maybe NOT meta-reality---but let's NOT go there.
This prediction is NOT falsified, and so the multiverse passes a non-trivial falsification test.
It doesn't make yours truly believe in the multiverse, but it does make yours truly take the multiverse seriously.