Yeah because it’s a certainty that will be implemented. No, that’s not how it works. Indexers will have to pay tens of thousands of dollars per month and I’m not exaggerating by any means, even with the current implementation of the protocol.
How is that possible? Well, it’s simple. Every single allocation costs ~$100 to open and another ~$100 to close, at ~150 gwei gas price. We will eventually have 50-100 subgraphs to allocate against. The default setup is having two parallel allocations.
Now let me translate that to you:
100 (usd, each allocation) x 2 (allo open, allo close) x 50 (subgraphs) x 2 (parallel allocations) = $20,000 per month
And this is a very conservative calculation I’ve done there. If you don’t allocate to most of the subgraphs, you’re losing a ton of rewards that come from inflation.
Now, changing the way delegation rewards work without overhauling the entire protocol design will most likely at least double that amount of money that we have to pay every single month.
Can you NOT put words in my mouth, maybe? And also maybe read the thread and conversations that went around that post before you point at someone?
Let me quote what I have said there, in case you missed it:
I think it would be better, if the OP idea is implemented […]
Now what was the OP idea?
According to this thread accounting for the time that a delegation was made when closing an allocation would allow other related changes:
- This would not allow “double counting” of delegation, which has the immediate impact of removing the need for a 28 day undelegation thawing period
- No “Delegator Front Running”, which has the immediate impact of being able to remove the delegation tax.
- More dynamic delegation market has implications for overdelegation, as savvy delegators can easily reallocate their resources without heavy taxes and waiting periods.
It seems to me that this would be more clean, fair, and dynamic design. The downside of course is gas efficiency. I think it would be worth exploring in this thread what that cost is and how it may be mitigated.
Would you be kind to tell me where in the world is it ever mentioned in the ORIGINAL POST of that thread, that delegators would be able to withdraw their rewards without going through the thawing period?
Even I mentioned the thawing period in the same literal post you quoted:
You will still have a thawing period for exiting the protocol, but you will have more freedom to move your delegation between the indexers.
Let me break this down for you, because I see there is a big lack of fundamental understanding in how the protocol works, and how you’re allowed to earn rewards.
Paragraph no.1: Yes, it’s completely fine for the indexers to do that, because they have much more risk and the protocol is NOT a one click sudo bash make-money.sh type of network where you sit on your sofa and you become a God.
It’s a full-time job that we have to take in order to ensure maximum efficiency, and it takes a lot of money, patience and efforts to reach that maximum efficiency.
Paragraph no.2: Without indexers you wouldn’t be able to get a single dime out of the indexing rewards nor the query fee rebates. 
If you think it’s unfair, why don’t you join the indexers side, yeah? It’s a permissionless protocol, you can take any side of the protocol you want. No one stops you from doing that. Grab a minimum of 100K GRT, a beefy infrastructure to support your stake, a few tens of thousands of dollars per month for operational expenses and let’s dance. 
Here’s my documentation for the Testnet, where you can learn how to operate before you deploy on mainnet:
Paragraph no.3: The infrastructure costs are nowhere near the operational cost that we have to pay in order to allocate. Read the beginning of this wall of text, I’ve outlined exactly what it costs us to operate efficiently. And this is not the only risk, we also have slashing events that can occur and we can lose everything that we have in a splitsecond, unlike delegators that will never be at any sort of risk.
Apart from what Fattox, Juan and Dave wrote, here’s some documentation you should read, as I see you definitely need it: