The distinction between "stress test", "spam attack", and "actual use" is kind of academic as long as the network can't handle either.
Why do you say it's misleadingly called a stress test? Regardless of motives it is effectively stressing the bitcoin network AFAIK.
"stress testing" and "maliciously stressing" aren't the same, motive is inherent in the word "test"
* The malleability attack: a transaction relayer is able to change the hash of the transactions, thus confusing senders or receivers who rely on this hash to check if the transaction has confirmed. This creates a nuisance, but all money arrive where they are supposed to. This attack does not affect memory usage on nodes, and it's an old and well-known issue.
* A transaction spam attack (misleadingly called a "stress test") where a shady group called coinwallet.eu creates a large amount of big transactions. These mostly have appopriate fees so that regular users who want their transactions to confirm in a timely manner has to out-bid the spammer. All the unconfirmed transactions are stored in the memory of nodes, so this severely affects memory usage (currently about 1 GB on some nodes).