There are essentially two categories of answer to your question: blockchain as a better ledger, and blockchain combined with physically unclonable functions to do the things you seek.
Blockchain As A Better Ledger
Blockchain can function as a better public ledger system than existing paper-based or even traditional, single source, trusted electronic counterparts, in many respects:
- It is distributed: Many separate servers maintain individual blockchain logs, ensuring that an outage by some few of them will not deny service to the remaining ones. You can count on some ledger being available for verification.
- It is hard to corrupt: Assuming that more than half of the blockchain servers are honest (and available), the logs among all honest servers will be identical.
- It is gaining popularity: Blockchain is growing in popularity due to cryptocurrency, and many companies are starting to consider adopting it for other purposes. Note, "starting" to consider - there are many issues including the fact that cryptography is hard for managers to get; plus, what is the business model for mining non-monetary transactions? What do I get out of it, other than "I mine your transactions, you mine mine?" Maybe that is enough.
Blockchain Combined with PUFs
Driving to the heart of your question, there is research into physical authentication systems that has led to something called physically unclonable functions, or PUFs. A PUF is a special function of a hardware device, such as a CPU or other computer chip (FPGAs, RAMs, etc.) that responds to a challenge in an instance-specific way. For instance, a ring oscillator can be created on the chip with a frequency that depends on both challenge bits given to it, and material variances in the fabrication process, and a counter can return the number of counts in a given amount of time. Since each device is different, it is hard for a forger to predict the number of loops for a particular challenge without having access---current or prior---to the genuine article.
For a device with a PUF, certain points in the supply chain could challenge the PUF and record the results in the ledger, and later, a recipient of the device could verify all the recorded challenges. There's probably a better way to do this without revealing the challenges, something where the points along the way could log a keyed hash of the challenge/response pairs using a key known only to the recipient.
For other physical entities, you could use a special chemical formulation that reacts differently to other "challenge" chemicals applied to it, so long as what is committed is not the actual response, but a proxy for it (hashed response). This would have the benefit of ensuring correct sequencing through the supply chain. If, at any point, someone replaced the genuine article with a counterfeit, they'd have to emulate the chemical state applied to the genuine article up to that point in the chain. Maybe I am overthinking it. Why not log hashed, very low dosages of radiation applied to a sensor securely tucked inside? Or the same with GPS coordinates of the item's path from shipper to receiver correlated to other distinct items couriered by each link in the shipping path? Etc.
Hope vs. Hype
You bring up an excellent observation, that much of this is hype designed to sell interest in a new technology. The biggest problem that I see is what's in it for me? To make blockchain work, each transaction submitted to the record must be "mined" by others in order to be added to the transaction chain. With cryptocurrency, the reward for successful mining is a slice of the coin. But how do you reward others for mining more mundane things like transactions?
You could create a company that does this "for a living," but recall that blockchain only makes sense when you have a large, non-colluding set of servers. There would have to be many companies that would agree to do this. Further, there would have to be compensation for logging items in the chain. As with typical service industries, there's a very real risk of many small startups being acquired and folded into a set of few megaproviders, and then they would be able to manipulate the official record.
To wrap up, I had to think about this question, as it's an important one. The key things that I would advise are that it's possible, but it's easier for some types of products than for others, and the big barrier is figuring out a reward system for those willing to maintain and commit to the ledger.
Hopefully, others in this thread can contribute to this discussion, or correct any errors I made.
Some further reading: