While scrolling, I found the data was lagging again. I thought it was a network issue, so I switched nodes, but it was the same. Later I realized the indexer was rate-limiting; the Subgraph's block sync was a bit slow, and the RPC was directly queuing you.

In short, on-chain data now also has tiered services. Free nodes are basically "take it or leave it." It's a lot like the logic of restaking—layers upon layers, sharing back and forth, and when something really goes wrong, you don't even know who to turn to.

I now make it a habit to have two or three sources on hand when checking things. If one gets stuck, I switch immediately—I'm not betting on a single point.