Using OH as master and phone as a slave is ideologically wrong. The device which is the "brains" of the system should be master. If the phone is slave it would always need a "master chip" at the other half, since almost all peripherals act only as a slave. Doesn't sound reasonable to me.

Multimaster solution would have a same problem, there is very little chips that can act as a multimaster. Whole multimaster thing is a little hackish IMO.

IMO phone being a master with interrupt line is the only reasonable solution, since most of i2c peripherals act only as a slave. With this solution there is whole lot of chips that can be connected directly to i2c and making other halfs is simple and fun.

Many MCUs can operate in all of the mentioned modes (master/slave/multimaster) so I'd guess phone SoCs can also do that. If the SoC handles all three modes it would be possible to leave out the interrupt line. I still think that if Jolla has done that, they have made a huge mistake.

Basically without knowing more it is not possible to choose what chip should be used. Though, no matter what the chip is going be the PCB will be about the same size, and changing the PCB for the new chip is quite simple and quick operation.

The hard part of the project is the mechanics. Since neither mechanical details aren't yet published, I don't know if there is much that can be done on that area either.

For now you can of course test different chips and their drivers, so that their strengths and weaknesses are known, when Jolla decides to publish the details.