That shouldn't be an error, I see no reason for the miner to check it. If the duplicate job was treated as any other job it would work.
Whether the pool shoud not send the same job twice is a trivial issue.

Edit: after further thought my opinion is the miner is at fault here. The issue should be raised with the xmrig devs.

When the miner starts there is a race condition between the miner threads and the stratum thread receiving the first job from the pool.
Sometimes the miner threads start hashing before the first job which could result in the miner crashing, submitting invalid hash based on
hashing random data, or it may go completely unoticed.

I have seen this problem and have fixed it in cpuminer-opt by adding a first job flag so the miner threads will wait until the first job is received
before they start hashing. I also saw this problem with verthash miner and open an issue https://github.com/CryptoGraphics/VerthashMiner/issues/32
but no action has been taken.

It seems kawpow may have tried to solve the issue by sending the first job automatically at login to speed things up. Resending it should not be a
problem. It's effectively a NULL change and a NULL change should never be an error.

Just my opinion.