Being a master is done by bit-banging on all PICs.

Multi-master I2C is much more complex.
It is a bus, not simple point-to-point.
Think of it as a tiny local area network.
You have similar issues.
As well as collision detection, arbitration, you must think
carefully about how your application deals with it.
Do you back off and retry endlessly?
When should you retry? Immediately? After random interval?
How often? Once? Endlessly?
If masters retry at the same time,
What is the volume of traffic?
This affects how likely these problems occur.

A vital question: Will any of your PICs be I2C slaves?
If so, the lack of I2C hardware will make this very difficult.
Even with the hardware, it's hard work.

What's your application & level of expertise?