Since the past month or so, this has been bugging us while trying to implement a RDM library. Initially we thought it was a bug in the code, but to my surprise, having dug up this thread .... coinicidences aren't that rare.
We have ended up with exactly similar situation(s), where collisions from Device A and B (having pretty close Device ID's) turn up as having good checksum !!
In our case in fact, the "collision packet" that does have a good checksum, decodes to a device id that doesn't even exist (but is close to the existing ones) ... thereby implying that somehow ... response packets from A and B come in at exactly the same timeframe on top of one another, forming this collison packet .
It'll be interseting to know if anybody else has replicated this scenario ?
It might turn out to be a shortcoming of the entire discovery process.
Cheers
www.enttec.com