If you do choose to use an oscilloscope with CAN decoding, I would suggest using two probes differentially and differencing them on the scope (usually summing the two channels plus inverting the low side probe). You may get a more reliable data stream that way, and avoid temporarily corrupting vehicle systems. The probe loading, as stated already by someone, is not a concern at all at CAN speeds. If you do it the other way, with the scope floating at CAN low line potential, you are adding massive unbalanced capacitance to the CAN pair, which despite appearances, are supposed to be balanced at AC. (Somewhat like a phone line, where there is about 50 V DC across the pair, but usually very good balance at AC to avoid mains frequency hum pickup).
On my most recent vehicles, access was/is possible via the OBD port and there are often separate high and low speed CAN busses there, one of them being on non-standard pins. It is fairly routine to modify the CAN to USB adaptor by fitting a double pole changeover switch to enable one or the other bus to be selected. Google may find the details for you.
I would venture to suggest that CAN is a mediocre interface at best, not helped by the liberties that vehicle manufacturers take with the wiring. Some use screened twin cable of the correct impedance, but have long gaps with screen pigtails at connectors, others think that they can get away with plain unscreened twisted pair, with very little care to keep the pair tight at connectors. At least one vehicle manufacturer does it fairly well, and their standards are available somewhere on the web. Others always claim to have followed the standard. I work for a major ECU manufacturer, and take great delight in ensuring that our specifications clearly state what is required of the CAN wiring. We test new designs properly for standards compliance too, while the vehicle wiring is, well, perhaps not tested at all. And don't get me going on the truly crude, vile and horrible LIN bus, or the extreme inadvisability of using the SENT interface for anything that needs to be reliable.
The ARINC 429 mentioned above is delightful, by comparison, and is only expensive to implement when you use the expensive chipset made by a certain semiconductor manufacturer. Most people use a small portion of their FPGA or DSP processor and a few op amps and comparators to make a cheap but still standards-compliant interface nowadays. But as stated, the external test equipment may well be hideously expensive. A small but enterprising and innovative manufacturer could easily corner the market with affordable equipment.