Lately, I've been seeing debates about "lag/ misleading tags" in on-chain data tools, and I'm not surprised... You think you're looking at real-time data, but in reality, there are several layers between you and the data: nodes, RPC, indexers. Some RPCs are queuing, some indexers haven't finished running, and even for the same transaction, you might see it confirmed on site A, but still appear unprocessed on site B. To put it simply, the "on-chain" data hasn't changed; what's changed is the pipeline you're accessing the chain through.

I now tend to treat on-chain observation like a thermometer: it can show trends, but don’t use it as a stopwatch. For important operations, I check two more sources, and if necessary, directly look up the raw transaction in the browser to avoid being misled by tags.

As for my understanding of "long-term"... it doesn't necessarily mean starting from a year. My own rhythm is more like about a quarter: enough time for the noise to settle, but not so long that errors drag on. Anyway, survive first, then compound.