According to CrpMapStat, Bentropolis has a 0.141% chance of flipping, Windurst has a 0.035% chance of flipping, and Pete has no chance of flipping. Assuming Swazi has a garrison of one unit, it presently has about a 1.3% chance of flipping. After either temple rush (after the first cultural expansion), Swazi would have about a 1.7% chance of flipping (it would have about a 2% chance of flipping after both rushes). Assuming Isandhlwana has a garrison of one unit, it has about a 0.13% chance of flipping, which would not change after rushing. If the temple at Pete was rushed, Windurst would have a less than 0.01% chance of flipping. If the temple at Bentropolis was rushed, it would have about a 0.09% chance of flipping. My numbers may be inaccurate due to having to estimate cultural difference and garrison of the enemy cities. The cultural difference is more significant than the garrison in my calculations. My numbers for our cities should be pretty accurate, my numbers for Swazi should be within 0.2%. I was probably slightly conservative with my estimate of our cultural superiority.
In addition, Windurst's flipping chance would be eliminated with a garrison of 2 and Bentropolis' flipping chance would be eliminated with a garrison of 4 (not that units would be well used up there). If Pete's garrison was reduced to 1, it would have a flipping chance of 0.05% without a rush, and a flipping chance of 0.01% with a rush in that city.
For people who don't want to figure out what that means, a rush in either city would create an equal flipping pressure on Swazi, a rush in Pete would practically eliminate flipping chances in Windurst, and a rush in Bentropolis would cut its flipping chance by a third.
If Swazi was not present, a rush in Pete would probably be better since would get 2 grassland, 2 sea, 1 coast, 1 hill, 1 flood plain, and 2 desert, while Bentropolis would get 1 plain, 2 hill, 1 mountain, and 3 desert. In addition, the happiness would help Pete more.
Edit: By the way, 3 squares gained through cultural expansion are significant for flipping; 2 would be gained immediately after rushing, 1 (not significant regarding our cities' flipping chances) would be gained after 10 turns.