Improved the way cached data is managed, reducing the amount of cached data stored over time.
The Budget/Stored Size row in the main window now shows the current stored size, refreshed while a backup is running and again when it finishes.
Google Cloud Storage: Arq now refreshes its access token before it expires instead of waiting for a request to fail. A token that expired in the middle of a large upload previously caused the whole upload to be retried, which could be time-consuming.
Arq Pro: computers managed by an Arq Pro group no longer show the "update available" notification when auto-update is off, since updates are controlled from the Arq Pro console.
Bug Fixes
Fixed the Arq7Agent service becoming unresponsive to the Service Control Manager (Windows event 7011) during a Remote Desktop logon. Arq7Agent asked Windows for device-arrival notifications for every device interface class, so RDP device redirection (the client's drives, printers, smart cards and ports) delivered a burst of notifications Arq had no use for. Arq now registers for volume notifications only; removable-drive detection is unchanged.
Fixed a failed read of the previous backup's directory data being reported as a backup error. That skipped thinning and budget enforcement and forced a full rescan on the next run, which could make a transient failure repeat indefinitely. Arq now logs the failure to the activity log and backs the directory up as new.
Fixed a race at the end of a backup where the pass that computes the progress-bar totals could still be reading backup records that retention thinning, budget enforcement, or unreferenced-data removal had just deleted. This also fixes the progress bar dropping to zero during cleanup on plans whose scan hadn't finished.
Fixed the emailed backup report showing 0 for scanned bytes and files when the scan-ahead pass hadn't finished.
Fixed a backup plan's "last backed up" date being updated by an aborted backup, which could silence the "no backups in N days" alert.
Fixed a backup plan's "last backed up" date never being updated when every backup was a back-up-and-validate run. This affected plans backed up to folder, SFTP, or pCloud destinations less often than every 30 days, such as rotating removable disks, and caused false "no backups in N days" alerts.
Fixed retry loops for Google Cloud Storage, Google token refresh, OneDrive, and pCloud that could retry a failing request indefinitely on certain network errors. Arq now gives up after a bounded time, while still waiting out genuine network outages.
