apples to apples
Zeev knows he's not comparing apples to apples (assuming when you refer to "apples" you're referring to data). he simply values use of a constant data SOURCE more than constant DATA. There are arguments to made either way, but I agree with you and others who favor focusing on keeping constancy of the data set rather than the source. It's great that the source with historical data available probably guarantees constant methodology and high confidence that what they measure, they measure with consistency. But if they're only measuring a fraction of the data output that they once did, the value is highly diminished, particularly if there is any self-selecting bias (as is likely the case here) in how the parties creating the data tend to choose the entity for collection.
If I was going to try and predict the outcome of an election and my choice was in doing that on the basis of data collected on only 15% of the electorate, but from a survey team for which I had data for a couple decades vs. on the basis of data collected on 99% of the electorate, but from a new source, provided each of the sources was reputable and using equivalent methodology, the decision isn't a close one.
From my perspective, the key here is whether the methodology used in collection and the integrity of the data of the "new" source can be assumed to be substantially the same confidence level as the former source.
Market Data 
Markets 
AI
