Excruciatingly slow ERC-20 token transfers export

[eresque]

I am using ethereum-etl to parse all specific token transfers from start to end block. As a node I'm running a local snap node with all the receipts being saved parameter (--txlookuplimit 0).

To test this instruments work I've decided to parse USDT token (0xdAC17F958D2ee523a2206206994597C13D831ec7) transfers from 20_500_000 block to 20_505_000 block.

The issue I'am having that it works really slow. I am talking taking about 15±0.5 seconds to parse all the transfers.
The official docs state that:

You can tune --batch-size, --max-workers for performance.

But tunning --max-workers parameters actually gives me a slight decrease in speed; --batch-size set to 4096 increases speed to 12.4±0.3 seconds, but it's further tweaking worsens the result.

For my original task I woult need to parse millisons of blocks, so the current speed is unacceptable.

The command I'm running:

ethereumetl export_token_transfers --start-block 20500000 --end-block 20505000 --provider-uri http://127.0.0.1:8545 --output token_transfers.csv  --tokens 0xdAC17F958D2ee523a2206206994597C13D831ec7

I've read this Medium article: How to Export the Entire Ethereum Blockchain to CSV in 2 hours for $10, but I can replicate, since I can't get access to AWS services.

Am I facing a hardware bottleneck or is there something I can do to make it faster?

Would much appreciate any suggestions!