A price delivered to a smart contract can be correctly formatted and faithfully recorded while describing an earlier market. Oracle review therefore needs a clock as well as a number.
An oracle connects external information to a contract
Ethereum’s documentation describes oracles as systems that make off-chain information available to smart contracts. It identifies correctness and availability among the central challenges. Recording an input on-chain does not independently establish that the external information is accurate or sufficiently current for every use. Ethereum documentation — Oracles
The relevant age starts before publication
Imagine a lending application using an invented collateral price of $100. The external observation occurred at 10:00, the oracle published it at 10:02, and a borrower’s transaction executes at 10:10. At execution, the observation is ten minutes old, even though its on-chain publication is eight minutes old.
If another timestamp records when a website retrieved that same value, it might make the display look newly updated without changing the age of the underlying observation. A review that checks only the latest retrieval time would miss the distinction.
Now assume the relevant market has moved to $85. A collateral calculation using $100 may overstate the contemporaneous value under these invented conditions. A different move could understate it. The issue is the mismatch between the data and the decision time, rather than a claim that every old price favours borrowers.
An update policy also needs context. Does publication occur periodically, after a specified deviation, or through another process? The number of reporting participants is useful information, but it does not answer how quickly a common stale input would be detected.
Fallback behaviour creates another decision. Rejecting stale data might stop an unsafe calculation while temporarily preventing a legitimate action. Continuing to use the last observation might maintain availability while increasing valuation uncertainty. Those consequences should be described rather than hidden behind an “oracle protected” label.
Inspect the data path end to end
Record the source observation time, publication time, update trigger and application’s permitted data age. Check units, decimals and the market represented by the feed; freshness cannot correct a price for the wrong asset or venue.
For a specific deployment, inspect what happens when the feed stops or sources disagree. WFN’s analytical question is whether the input remains appropriate for the action being taken. The existence of an oracle is only the start of that review.