A commercial package no longer re-fails every boot
When a deployment's default-install set swept in a commercial package — one with a price or a sales contact — every restart tried to install it, was refused (correctly: an unattended boot has no one to pay or approve), logged the refusal as an error, recorded the package as FAILED, and promised "the next boot re-attempts it; that retry is the repair". For an authorization refusal that promise was false: no retry can conjure an approver, so the same errors re-appeared at every boot of every instance, forever.
The installer now classifies that refusal for what it is — terminal for an unattended installer — before attempting anything:
- The package is recorded as skipped, with its reason, on the install ledger
(
Plugins/_DefaultInstallLedger) — one warning naming the packages and the remedy, instead of a per-boot error with a stack trace per package. - Nothing re-attempts it. What changes the outcome is an event the next pass observes, not a retry — a Global Admin installs the package from the catalog, or the package stops being commercial. The classification is re-derived from the current catalog on every pass, so the moment a package's terms change it installs with no further ceremony.
- A genuine install failure (a timeout, a malformed package) keeps its existing behaviour: recorded as FAILED and retried next boot, because there a retry really is the repair.
If your instance should carry a commercial package, the ledger now tells you exactly that — and that installing it from the catalog as a Global Admin is the way to get it.