There has been a problem with page parts in a table that is a particular annoyance: it is an audit log of the activities in the database that is over 1TB in size. The main indices refer to record type NVARCHAR (100), as this is far more useful for five record types than a tinyint and record ID, which is instead a NVARCHAR (200) of the record's integer key.
They also cover indices, where the key, the old value, the new value, etc. are very wide.
This is an old system, and unfortunately the code for this check is not centralized in one procedure, but everywhere. It can not be changed and we will go through the painful process of a long rewrite of microservices.
So I reduced the fill factor for two of the indices from 100% to 85%.
And the side parts have gotten worse. I would say about 3x more side panels.
Is that a common result? Most recommendations suggest that the fill factor should be reduced to improve page splitting performance. I can understand why this is possible because the data in the keys are so wide.
Would it be advisable to reduce the fill factor further or reset it to the original value?