Because of wrong implementation of file locks in popular remote filesystems.
Storagenode uses sqlite for databases, they are not compatible with the network filesystem protocols. The storagenode also need normal locks for pieces too, thus the main storage is not compatible with the network filesystem protocols too.
There are some exceptions - like use the SMB when both - server and client are Windows. Unfortunately it doesn’t work on Linux neither their implementation of SMB, nor NFS.
So, in general if you forced to use a network access to your storage, then you can use iSCSI, it’s block-level network protocol and you would have a normal filesystem there like ext4 for Linux or NTFS for Windows and it’s can be used for storage. However, if you would have interruptions in the network you could have missed or broken pieces and as result your node would have a greater probability to fail audits and being disqualified at some point.
We have also a positive reports from operators in using Ceph: Distributed architecture for SNOs for high availability of storage nodes - #7 by Cmdrd