Lately, people keep asking me: why do I see that some transaction has “already succeeded on-chain,” but you still don’t have a trace on your side? Straight to the point—what you’re seeing isn’t really “on-chain.” It’s the version your connected node/RPC/indexer is giving you. The node might be a few hundred slots behind; if the RPC load is high, it may hand you cached old results; and as for the indexer—forget it. Replay it once and it can reshuffle the order so it looks like time travel.

That’s also why, when cross-chain bridges go wrong and oracles throw errors, everyone suddenly collectively “waits for confirmation.” It’s not panic—it’s because they know the final state you think you’re getting may just be an illusion produced by some service provider. And I’ll add one thing: what I fear most isn’t slowness, it’s chaos. If it’s slow, I can wait; if it gets chaotic, you won’t even know which data source to trust. The more you look, the more anxious you get. Anyway, I’m used to checking two or three RPCs at the same time, and I also add a lightweight node I run myself to compare—better to be a bit more troublesome.