I have the impression blockchain reorgs are more prevalent on testnet?

Question

They also seem more significant if they happen, e.g. I just see REORGANIZE on height: 839127 of 839203 in the log file, which is quite a number of blocks. Can anyone comment whether these observations are a correct / typical (or an exception)?

If that is so, why is testnet more "vulnerable"? Is it because the network's hash rate is massively lower on testnet vs. mainnet? Intuitively, I would have thought it is simpler to agree on the longest chain if there are fewer nodes, but maybe that is flawed logic from my side?

FYI: I have read this explanation already, but wondering what the differences are with respect to testnet vs. mainnet.

Answer 1

In theory, yes, fewer nodes should make things easier to agree upon, but in practice, that's not always the case. A few things might contribute to a big reorg on testnet that wouldn't usually happen on mainnet:

1. The testnet's hashrate is massively lower than the mainnet. As in, I could attack the network with my personal computers alone and have way more than 51% most of the time.
2. As a result of 1., new person joining the network might bring enough hashrate to accidentally "attack" the chain, meaning they might start mining on a different top block than the rest of the chain and cause confusion on the network for a while.
3. Alternatively, someone might intentionally attack the testnet just to try out things or test for robustness.
4. Most likely, nodes on testnet are not optimized for connectivity, etc. I doubt there are many nodes running on cloud services like AWS for instance. As a result, the latency problem described in the linked post is likely exacerbated, and so the few miners on the testnet are probably more likely to end up on different chains.
5. Building on 5, lets say there are basically 2-3 people mining on testnet with roughly the same hashrate. If they end up on different chains, it's decently likely that they will occasionally diverge on what they think is the longest chain and continue to disagree for some time, as both chains might keep up with each other within the latency lag window described. More miners would actually help to mitigate that issue, as blocks would be propagated from more locations around the world, and thus have more variance in latency between miners.

There are probably some other reasons as well, but these are the ones that come to mind right away.