It would be easier if all attributes get emitted under a single event. If that's too much of a headache than we should at least add the DR_ID as an attribute so we can link the event to the relevant DR.

DistributionsFromGasCalculation() constructs a series of event attributes and emits them at once at the end. Since an attribute is just a key-value pair, proxy & executor gas attributes' values follow the "ID,amount" pattern.

The execution needs to be aware of the base gas fee that was already consumed.

Shouldn't the base gas be taken from the execution gas limit while the tally program uses the tally gas limit?

No, the base gas is charged from the tally gas limit. If we were to take it from the execution gas limit all the overlay nodes would also have to consider the base gas cost when calculating their individual limits. And we said it is to cover some of the computational load a DR causes for a validator that are not charged directly like the filter and tally VM; think basic consensus check, median gas price calculation, etc.

OK added. Perhaps BuildFilter() should check if the filter gas cost is within the tally limit?

gasUsed is not guaranteed to fit in a uint64 right? Lets say I change the price of my data proxy to 1 billion SEDA from 1 attoseda. Due to the delayed update the overlay computes its gas cost using the 1 attoseda fee but when the tally checks the price it has flipped and is now 1 billion SEDA.

Obviously an extreme example but think it's worth considering.

where is this used?

We charge the base gas cost for every data request.

When the request times out or when there is no basic consensus, we charge the commit cost for each executor that did commit (CalculateCommitterPayouts())

In terms of events I think we're only really interested in 2 values: medianGas and lowestReporter (which can be empty for uniform scenarios). This will allow us to simple add those attributes to the DR completion event and simplify indexing in that regard. The adjusted values after the shares calculation we will receive from the contract when it sends the actual SEDA amounts to the identities.

Please take a look at the function DistributionsFromGasCalculation() and let me know what you think. If we want those values, we can add those values to the GasCalculation struct and emit them as events later, too.

Can we not add the RF=1 case to areGasReportsUniform? Since we check for 0 there as well.

Removed the replication factor check and changed the check in areGasReportsUniform to

if len(reports) <= 1 {
  return true
}