Time Machine backups are very slow due to Outlook's file-distribution algorithm
Outlook stores new message or attachment files evenly across 256 different folders in the profile directory. This even-distribution algorithm unnecessarily slows down Time Machine backups and also hurts battery life. If instead Outlook stored messages and attachments in "batches", e.g. first 100 files in one folder, next 100 files in the next folder, and so on, then its Time Machine impact could likely be reduced over 90%. This could be a big improvement in backup times and battery usage for users with large mailboxes. Note that users with large mailboxes are often IT Pros and IT decisionmakers who control whether a company will purchase and use Outlook!
Here’s more info: When Time Machine backs up a folder where none of the files in the folder have changed, then TM creates a folder on the backup drive as a hard link to the previous backup’s folder. Only one hard link is created, regardless of the number of files in the folder. This is very efficient. However, if any files have changed in a folder, then TM needs to create new hard links on the backup drive for every file in the folder. Furthermore, in addition to just creating the link, TM also needs to read file attributes and extended attributes of each file. While both link creation and attribute reading are fast (total of 0.05-0.10 seconds per file on my MacBook Pro), if there are 1,000 files in a folder then each folder will take TM 15-60 seconds to back up the unchanged files. Multiply that by 256 message folders and 256 attachment folders, and you end up with "hourly" backups that take 30+ minutes, where most of that time is creating links to files that have not changed! All that I/O also kills battery life.
The main thing that Outlook could do to reduce its impact on Time Machine backup time and battery usage is to reduce the number of message/attachment folders that have any file changes. If a user gets 100 new messages and those messages are put into 100 folders, then Outlook needs to create new hard links for every unchanged file in each of those 100 folders. If instead those 100 new messages were put into only one folder, then Outlook would only need to create hard links for unchanged files in that one folder. I’m not saying that Outlook shouldn’t distribute files across folders—only that if you changed the distribution algorithm to “batch” the updates, you could have a vastly smaller impact on the backup times and battery usage of Time Machine.
I’m not sure whether this kind of “batched” distribution would be an easy change—your engineers will know—but its impact on the usability and performance of Macs running Outlook could be a huge win, especially for users with very large mailboxes.
If it matters, senior IT decisionmakers (the people who decide whether to use Outlook across a large company!) often have huge mailboxes! So this would be a shrewd business decision to make for users who have an outsized impact on Office revenue.