A little hyperbole going on here. As noted in the article, Bitcoin block 666833 produced two competing chains. This happens from time to time. One of the chains was mined by SlushPool, the other mined by F2Pool. The SlushPool mined transaction confirmed spending 0.00014499 BTC to address 1D6aebVY5DbS1v7rNTnX2xeYcfWM3os1va. The F2Pool mined transaction confirmed spending 0.00062063 BTC to the same address.
After a few more blocks, the network agreed that the SlushPool mined block 666833 was the correct one, invalidating the transaction mined by F2Pool in their 666833 block.
Thus, for a brief while, the same coins were spent twice. But, as designed, the longest chain won by network-wide consensus, thereby invalidating the double spent funds on the “loser” chain. This happened within the next 6 blocks: the typically accepted number of blocks allowed to secure/guarantee a Bitcoin transaction from double-spends.
The other curiosity was this double-spend transaction was not the result of a RBF transaction, as this is the most common cause for many double-spend transactions. A RBF (replace-by-fee transaction) is by definition a double-spend. A RBF transaction will have a higher miner fee and is intended to be mined ahead of an earlier transaction in the mempool spending the same UTXOs by virtue of the higher miner fee. Network consensus will then prevent the earlier transaction from being included in a later block since the UTXOs are now spent from the RBF transaction.
