I would argue that this is a given – the DAO should absolutely refund gas costs to all oracle operators (and for expenses already incurred this should happen ASAP) and it should put in place a process so that these costs are either covered up-front based on a best-effort guess (+buffer), or repaid on a timely schedule (e.g. maximum once per month). Both could be accomplished in a clear and simple manner via something like Easy track.
Oracle operations should be fully transparent and included in the calculations of determining protocol profitability etc, so the closer this is tied to on-chain ops/governance the better.
On the larger initiative proposed by @eliasimos I fully agree and think it would be great to increase the oracle set, although based on Elias’ calculations it seems that the overall cost increases quite considerably as we add more members. Would something like rotating the oracles work? E.g. if we grow to a set of 11 but for each week there is an “active” set of 8 (7 for consensus and 1 backup) oracles that are “on duty” for that week?