Ok Derek I hit the bone yard today and cut a couple CAN C and CAN B bus bar/hubs and leads out of a late model E Class for testing with. Resolved my flaky connection issues, but it didn't resolve the baud rate issues as I hoped.
I had actually loaded a custom sketch and custom CANbus.h hacked together with some pieces i found here.
I also set all the CAN transceivers to (LISTEN) only rather than Normal. On first connection it stated capturing data, but would stop after about a minute. Couldn't use the Nodejs features either, but did log some with the USBPcap. Thinking having a good connection fixed it, I loaded the CBT autobaud sketch to use the Nodejs features. Nope bad move, it set me back to square one. And then I realized I didn't save my previous test config....
So starting over from scratch, here is what I found today. Maybe it will help us nail down the 83.3k network issue.
When the MB CAN B bus is in sleep mode, testing between CAN high and low registers 11.7 volts. Soon as you trigger a wake up event, that switches to 3.5v.
When running the the autobaud sketch, or any sketch with transceivers set in (Normal) mode, the CAN B bus will not wake up. If you wake the bus, and then connect the CBT, it crashes every node and causes all sorts of funky actions in the car. Headlamps come on, command unit malfunctions, and red "Visit Workshop" message in DIC. Like a voltage short issue. Same thing happens if you bring it out of sleep mode by turning the key on. Doesn't happen when CBT is in listen mode.
Should also be noted that the CBT is locking up under same events even in listen mode. If bus in sleep, CBT acts normal, no errors. Soon as the bus wakes up, CBT will capture a single wakeup message, then locks up and has to be power cycled. Some times the Red LED will come on depending on baud settings.
All these tests were done with only USB powering the CBT. No other connections besides the CAN B network. I'll try to rebuild the working sketch tomorrow.
Hope this helps, I am dying to get ahead of this.