[BUG] Incorrect transaction count in gas.usage_by_address model

[sohjag]

Description

The gas.usage_by_address model is significantly undercounting the number of transactions for addresses.

Current behavior

For example, for address 0xC94eBB328aC25b95DB0E0AA968371885Fa516215:

• gas.usage_by_address shows 20,534 transactions (query)

SELECT number_of_txs
FROM gas.usage_by_address 
WHERE address = 0xC94eBB328aC25b95DB0E0AA968371885Fa516215
AND blockchain='ethereum'

• gas.fees (source table) shows 1,397,874 transactions (query)

SELECT COUNT(*) as tx_count
FROM gas.fees 
WHERE tx_from=0xC94eBB328aC25b95DB0E0AA968371885Fa516215
AND blockchain='ethereum'

Expected behavior

Transaction count in gas.usage_by_address should match the count in gas.fees table where currency_symbol is not null.

Impacted model(s)

gas.usage_by_address

Possible solution

Investigate incremental processing logic in the model. May be a full refresh or modification of the incremental merge strategy to properly aggregate historical transactions?

[pfedan]

I just ran the spellbook as a normal query, and it returns correct values: https://dune.com/queries/4768584

This suggests that the initialization has not been run correctly, or that the incremental logic isn't working as expected.

[jeff-dude]

tied PR to issue