Recently, someone asked me: Why does on-chain data keep "stuttering," even though blocks are being produced?

Honestly, it's not that the chain is slow; it's the layer you're viewing that's gasping for air: Subgraph needs to wait for the indexer to finish scanning events, write to the database, and then you can query;
On the RPC side, if there's rate limiting or queuing, your request is like squeezing into a subway, waiting behind someone else.
Plus, some front-end applications use public RPC as a fallback, switching back and forth, with timestamps jumping around, making it look like "frame drops."

Why do I see different order book depth and trades before and after executing a trade?
Because what you see might be cached or outdated index data; at the moment of the trade, it's the real on-chain data plus MEV front-running.

Recently, during extreme funding rate swings, this is even more obvious.
People argue whether to reverse or keep pumping the bubble; I just think: when the market heats up, query volume surges, RPCs are more prone to glitches.
Don’t interpret "stuttering" as a market signal…
Switch to a self-hosted or paid RPC, or wait for an extra confirmation before trusting the data, to avoid blaming slippage again.