One (granted, fairly minor) issue I encountered lately was a "tx-size-small (code 64)" error when broadcasting a transaction to discard a small UTXO. The most efficient way to do this that I could see was to simply create a transaction with a single vout consisting of an empty OP_RETURN, but it turns out that such a transaction, while recognized as valid by decoderawtransaction, is below the MIN_STANDARD_TX_NONWITNESS_SIZE defined in policy.h and is therefore rejected for relaying.
You can of course just add ~30 bytes of junk data to OP_RETURN to reach the size limit, which makes the transaction relayable and mineable with the current policy. Similarly, in the case of any other otherwise-standard transaction that is valid for relaying outside of the limit defined in MIN_STANDARD_TX_NONWITNESS_SIZE , you could just pad the transaction with another OP_RETURN or similar vout. Therefore, this restriction seems both rather pointless and wasteful.
There are some valid reasons why you might want to junk a UTXO, such as if you do not want to combine it with other UTXOs for privacy reasons (say in the case of dust being sent to an address by a third party for de-anonymization purposes), or if the wallet is empty outside of said UTXO and it would be below the dust limit if you sent it elsewhere. You could of course just leave it, but many wallets do not have coin control or coin locking facilities, which means it might inadvertently be included in a transaction at a later point. As such, while I recognize that this is somewhat of an edge case, seeing as the size restriction does not really seem to fulfill its stated purpose of discouraging non-standard transactions anyway, would it not make sense to loosen this restriction to allow otherwise-standard transactions with just one single-byte vout?
