Please rethink webhook logging changes

A change was recently made to the way webhook logging works, which has broken a common use case. If a webhook fails for whatever reason, it used to be possible to go to the logs and resend the request - but now that webhook logging turns itself off after 12 hours, you're unlikely to catch a failed request in the logs (unless you manually turn the logging on again for every repo every day when you get to work, so that you have logging enabled throughout the day... but ain't nobody got time for that). See #17498 and the comments therein for examples of other people frustrated by this change for the same reason.

If storage space was becoming an issue with the previous approach, perhaps a more useful compromise would be to impose a rolling retention limit - either storing logs for only X hours, or storing only the latest X requests per repository. Perhaps there could still be an option (for those who need it) to temporarily enable full log retention for a longer period (i.e. if the standard rolling retention period is 1 hour, the option could still be to retain logs for up to 12 hours, per the current option).

Note: I have filed this as a bug rather than an enhancement, as this is effectively a regression (a use case which was previously supported is no longer supported after the change to webhook logging).