FAT32 was "essential" for Acronis partition image backup files.
My experience a few years ago was that when restoring the system partition the Linux drivers took 3 times as long if the backup was on NTFS partitions
(I had one USB2 drive with both NTFS and FAT32 partitions).
I am happy that Macrium with WinPE is equally fast restoring from either.
When adding a new partition in unallocated free space on my internal drive something conflicted with the partition manager at the final stage of "APPLY".
I did not get my new partition at that point in time.
I got nothing - totally dead - C:\ was missing.
Plugged in my Partition Wizard Boot CD and saw all my partitions were missing.
Boot CD included a Recovery Wizard which saw all the boundaries of all the partitions I had used and recovered them.
Failed to boot but the Boot CD then fixed the MBR and I booted into C:\
Booted up and was instantly told one partition was corrupt and needed chkdsk which I did.
Lots of errors were fixed by chkdsk.
4 out of 4 NTFS partitions went missing and after recovery their files were intact and chkdsk was not suggested.
1 out of 2 FAT32 partitions were badly corrupt after the same experience - and I had no confidence in the quality of the "survivor"
Fortunately I had backup images on an external drive of all the important partitions.
I disliked FAT32 for the security restrictions, but decided that NTFS seems better able to survive disaster.
One other benefit to NTFS is that "Everything" will search the MFT's and in less than 2 seconds locate all files that match my query,
where a Windows search would take an extended coffee break.
FAT32 has no MFT so Everything does not look there.
I believe there are other super quick tools that may similarly only work on NTFS and not FAT32.
If you convert, please be aware that it is better to delete and then create a new partition.
I believe NTFS sector clusters should start at a different offset from the start of the disk than FAT32,
and if you preserve the files by converting FAT32 to NTFS then the start offset is wrong and the result is sub-optimal clusters.