However, it's possible that deteriorating conditions could cause the leakage on each phase of a 2-pole GFCI breaker to slowly creep up over time so that the leakages match each other closely enough that the breaker doesn't trip. These leakages could be from L1 to L2 through the pool. Or one phase could have leakage to the EGC or the earth, and the other phase have leakage to the pool.
I'm thinking about this.
This isn't the same as a balanced MWBC. If the water becomes energized,
even if the two leakages are equal, there should still be current flowing into the earth that is not returning on the neutral. I think it would behave like two individual circuits.
...
I agree with your statement in bold above. However, to illustrate a point consider the following examples of leakage at outputs L1 and L2 of a 2-pole GFCI breaker with a load neutral N:
1. 12 mA from L1 to pool water, 9 mA from L2 to pool water.
2. 12 mA from L1 to pool water, 9 mA from L2 to EGC
3. 12 mA from L1 to pool water, 9 mA from L2 to earth
4. 12 mA from L1 to earth, 9 mA from L2 to earth
In none of these cases is any leakage current flowing through the load neutral.
The leakage current from L1 to the pool water could return on L2, the EGC, earth and GES, or a combination of thereof.
The leakage current from L2 to the pool water could return on L1, the EGC, earth and GES, or a combination of thereof.
But in all of these cases the amount of common mode current through L1, L2, and N (i.e., the RMS value of their vector sum) is 12 mA - 9 mA = 3 mA, and this is what the current transformer inside of the GFCI responds to. Therefore the GFCI will not trip while under these leakage conditions.
It is true that any current flowing out of the GFCI (including any leakage) must match the current returning to its output terminals within less than 6 mA , or the GFCI will trip ( 4 mA or less to guarantee that it won't trip) . The neutral is not necessarily involved in this if there's a 2-pole breaker, but it would be involved if there was a ground fault on the neutral.
For simplicity I'm assuming that all leakage currents are in-phase with their associated applied voltage, or they're at the same phase shift.