Has there ever been a case where this occurred (with a market clearing mining fee is included)? You only need one miner to include the transaction.
Any nonstandard transaction will be rejected by 90 if not 100%, miners, but if you get lucky to get someone to mine it, it will work fine with the protocol.
The nonstandard txs are any that are something more than a transfer, multisig (i.e. the ones that have some sort of smart contract capability).
This is not theoretical - literally killed my project of decentralised oracles (Orisi) back in 2014. Our beta transactions were accepted only by 10% of miners, so you had to wait for hours to get your tx through. Perhaps it changed - I think in 2018 not even one mining pool accepted that, but perhaps by now they fixed it, although I doubt that.
May seem like a small thing, but I met other projects that bumped into it. They had the tech, just didn’t manage to convinve miners to accept their transactions.
You are incorrect about what constitutes a standard.
The "smart contracts" by their own are standard - the whole Lightning Network is based on smart contracts - as well as many other protocols running live for years.
It sounds more like your specific transactions were non-standard due to excessive size, usage of non-standard opcodes, or other technical quirks.
But this was 2014, so it may have been before your time.
*. https://www.clearytradewatch.com/2018/12/ofac-lists-digital-...
You are correct that you only need one miner, but that miner has to win a block.