It is common when analyzing paired-end whole-genome shotgun sequencing data to check for and eliminate PCR duplicates. The reasoning is that the probability of sampling a fragment of the genome of the same exact length from the same exact position in the genome multiple times is extremely low. Therefore, if you see multiple read pairs with the same exact sequence in both pairs, these don't represent independent events but are multiple copies of the same event.
My question is whether PCR duplicates can be reverse complements of each other. All the PCR duplicates I've observed myself have been exactly the same sequence on exactly the same strand. Is there a technical reason why this is the case, or is it possible that PCR duplicates could not be on the same strand as each other?