I dont actually see a problem with it as long as you dont run into a bus contention situation. I would use a third pin between the two, make one pin an input and an output. Input when not wanting the bus, output when you do want it.
If one chip wantes to use the bus, it first checks to see if this line is high. If it is, it skips on to it's next proces, or waits a reasonable amount of time before it checks the pin again. As long as this pin is high, it is not allowed to use the bus.
If the pin is low, it then makes it high. This means that it has taken control of the bus, and lets the other picaxe know to keep off of it.
For normal operations, both picaxes should have thier "contention resolution" pin, or whatever you want to call it, will be an input. It will only become an output when it's ready to take control of the I2C bus. The I2C pins are otherwise ignored completely.
Upon completion of it's I2C duties, it lowers the contention pin, and returns it to an input state, and goes about it's business.
There are also I2C multiplexers, for doing jsut this sort of thing. They're alot more expensive than a piece of wire and a few lines of code though. They handle the contention, buffer the I/O and everything on thier own so you dont have to worry about it.
You may also run into capacitance problems doing it the way I described, depending on how many chips you use, etc. Unless you can tri-state the picaxe pins (go into Hi-Z) which I dont think you can, it's capaitance will ahve to be added to the total in the line, and cannot psh it over 400pf or youll have timing issues.