In the random nature scenario, probably 99% of all the instances would be immediately discarded for potential because of 1 or more clearly suboptimal conditions.
Let’s say the labs are 99.999999 “efficient” compared to the oceans. That carved off 8 decades, meaning Earth’s experiment was still a million million million times bigger.
Let’s say I have a huge lab and we have 1,000 one litre experiments running continuously. It would still take me 274 years going non stop to recreate the conditions.
If it really was likely to happen in a few litres of chemicals over a few years, once those conditions arose in the early oceans we'd expect it to have happened trillions of trillions of times, not just once or a few times.
On the other hand if the conditions were so rare that they only occurred once in a few litres of ocean for a few days, those conditions must be so specific and unusual that we might never figure out how to create them.
You see we’re still figuring out what the conditions should be. Yes we’re good and recreating specific conditions, but we’re still figuring which ones we should strive for.
Natural of course had the luxury of just trying them, in parallel, over vast amounts of time.