Here is the latest test I tried, which makes no sense at all.

I send one tx at 1.0 sat/vB. Somehow, that is the only transaction that sent with the fee I agree to pay.
Here is the mempool.space link to it:
https://mempool.space/tx/eb813d7d4ed2973cce907ee6e6055bc3d70498e37d67d17885d2b9604eabdd0d?mode=details

Than I sent a second on bactched with the first one. The specified miner fee I chose was 0.1 sat/vB, which normally would be rejected because it's below 1 sat/vB.

But it didn't reject it, and it also didn't send at all at my specified sat/vB. Instead, it doubled the miner fee for the batched tx's at 2.01 sat/vB. Which is essentially more exoensive than sending separately, without batching.

Seperately they would have been 0.15$ each. But batched, they were 0.36$. So at this point, in thus instance, RBF/batching is purely harmful.

So I decided to add an other one to the batch. I decided on a 1sat/vB fee for that third one. But dumb Electrum basically increased the miner fee from 2.01 sat/vB on this one. Basically not increasing the fee at a meaningful level.

So when I input 0.1 sat/vB, it doubles the fee from 1 to 2 sat/VB.
And when I input 1 sat/VB, it barely increases by only 0.01 sat/VB.

I don't like this one bit. I want the old Electrum I was using a year or two ago. This is regression.
/