At AboutUs we hosted a Mediawiki as "about pages" for all sites on the internet. At times we wished to make changes to markup to, for example, improve how our users would enter addresses.
I used ad-hoc PEG parsing to find candidates for improvement. This had a short cut for testing but would run for many hours finding candidates for improvement. See Exploratory Parsing
Support and I produced a CSV with exactly what was found and exactly the recommended improvement. This was turned over to other engineers that applied the changes at a rate that wouldn't overrun read-heavy databases.
Most edits would apply but some would be blocked by concurrent edits by end users. Blocked edits were discarded to be discovered in a later pass.
digraph { rankdir=LR node [shape=box style=filled] node [fillcolor=palegreen] mgmt [label="Management\nInitiative"] com [label="Customer\nBehavior"] supp [label="Support\nAwareness"] eng [label="Engineering\nResponsibility"] arch [label="Architecural\nOpportunity"] mgmt -> supp -> arch -> eng -> supp arch -> mgmt mgmt -> com -> supp }
Initiatives would stretch out over days or weeks of casual cooperation producing work products that could be reviewed and applied at the next stage without the need for any step to be exceptionally clearly defined and executed without error.