Libra has a different value proposition than Ethereum because it's not decentralized and can't function as a global smart contract platform.

There's no way Libra Association members are going to ignore an order from an American (and likely European) court, and especially not a direct OFAC Executive Order. The link leads to sanctions against two Iranian bitcoin addresses - if they're willing to sanction bitcoin, there's no way in hell Libra gets a free pass.

A good example of Libra's vulnerability to the American government is SWIFT:
SWIFT is a cooperative society under Belgian law owned by its member financial institutions with offices around the world. Very similar to Libra associaton.
What happened? They banned Iran because of American sanctions.
America is also threatening Russia with a SWIFT ban.

Imagine a Russian company buying oil from an Iranian company, with the help of Libra smart contracts. Then an European company in turn buying something from the Russian company, whose products are in turn used by an American company.
The inevitable OFAC order happens and both the Russian and Iranian companies are banned, their on-chain assets confiscated. This in turns hurts the European company, which in turn hurts the American company.

For this reason I'm very skeptical of even American companies moving their business processes to Libra smart contracts, not to mention other countries. It would be SWIFT x100, as in a smart contract economy a ban would mean stopping all business operations, not just the main avenue of payments, which is more easily circumvented.
Even Germany is unhappy about reliance on platforms de facto controlled by the American government.

Ok, but why is this good for ethereum?

Because these issues are also true for almost all "ethereum killers". All DPoS and PoA networks, and/or those with governance, have the exact same problems (at best, they can move their reliance to China or Russia), while being worse in literally every other aspect: decentralization (in terms of vulnerability to validators forming cartels and harming the network for individual gain), adoption, likely quality of software and scalability (each Libra node is going to cost hundreds of thousands to run).

Networks that are architecturally decentralized but with an extremely centralized token distribution - eg. Algorand reserved so much stake for itself it has to promise to not stake more than 49% at the same time - are in this set too. The best PoS algorithm in the world can only make the network as decentralized as its token distribution. Libra's limit of maximum 1% per validator is more decentralized than this.

The answer to a question: in what aspects is Ethereum better than Libra is decentralization and immutability.

The answer to a question: in what aspects are platforms mentioned previously better than Libra - is nothing. Libra is better in every single aspect, they have no reason to exist.