one small integer is not enough, two small integers are better (lab journal initials/number and sample number, like AC7-295. something like AC7-295A, then AC7-295B and so on if needed. that's how i do it anyway) this way there's no possibility of mixup with other people's samples and samples described in old lab journals
After much debate over copious drinks at the bar, we finally decided to settle the argument with darts.
0 are all crossed.
1's are written as l
7's are all crossed.
And 9's... Well we got kicked out and it was never settled. How was I supposed to know the nickname Nicky sounds like the French word "Niquer" and somebody (Nicolas) got all bent out of shape over it. "Hey Nicky it's your turn!" apparently was not well received by a drunk frenchman.
But if they do get jumbled, sorting them back out into different experiments, batches or subjects or time periods might make you prefer some extra info accesible by eye.
If you've got a robot sorter maybe a qr code - but you'd have to be pretty large scale for that to be cheaper than a human.