We typically do a lot of reporting out of the changelogs to properly capture the number of defects opened against a release and to categorize their final status. It is particularly important for us to capture defects that are deferred from one release to the next. Reporting against the Jiraissue table causes deferred defects to simply drop off of the face of the earth.

It would be more convenient (and moderately more performant) to do these types of reports if a defect's complete lifecycle were captured in the changelog rather than everything that happened after the issue was originally entered. Please note that this would need to apply to any kind of issue creation event (user created, import created, SOAP created, e-mail handler created).

For my purposes, all I really care about is capturing the creation event. I don't care about capturing the initial state of each of the fields. If I really need to get there and because this is almost never an issue, I can infer that by working backwards from the current state of an issue using the changelog.