Another example from a favoured database - MongoDB. Rather than trying to come up with optimal query plans they try all pertinent ones concurrently. Whichever

Another example from a favoured database - MongoDB. Rather than trying to come up with optimal query plans they try all pertinent ones concurrently. Whichever finishes first is noted and used in the future. Its performance results continue to be monitored and if they start diverging all are tried again. This means not having to guess big O correctly, or access patterns or other stuff in advance. It is possible to add future optimisations and not worry about current ones since they just won't be used work inapplicable workloads.