• Eth POS takes roughly 15 mins to finalise, which is roughly equivalent to 2 epochs (64 slots). This is too long and bad for user experience. This fork choice rule + finality gadget also creates protocol complexity, increases the attack surface and needs hundreds of lines of code for things like validator shuffling every slot
• Finality Today - Every epoch (32 slots, or 6.4 min), all active validators get a chance to attest once. After two epochs, the Casper FFG finality gadget finalizes the block, after which point reverting that block requires at least 1/3 of all validators to burn their deposits
• Signatures of validators in committees are aggregated by a designated validator using BLS signature aggregation such that verifying the combined signature requires only one extra elliptic curve addition (not multiplication) per participant, and an arbitrarily large signature fits into 64 bytes, requiring only 1 extra bit per participant to store who participated.

Improvements in BLS aggregation would soon allow full on traditional BFT consensus which allows finalising a block before starting a new one (Single slot finality) without changing the hardware requirements for the validator

There are many considerations and design choices to consider for Single slot finality -

References

https://notes.ethereum.org/@vbuterin/single_slot_finality