So victims can only make sure that they have a malware checker that finds the culprit, then do fresh installs, then check each file before it's restored from backup. Sounds like a crazy amount of work.
The attacker could: get the victim infested with malware; ensure that the malware infects all new files; wait one year; then trigger the encrypt function.
If the victim pulls a file that is newer than one year old from backup, the victim will pull the malware from backup.
(Only if the file is of a type that could be infected in the first place, of course. README.txt will not be infected.)
For instance: I have backups going to an append-only s3 bucket in a separate AWS account, but I don't have monitoring in place to ensure that bucket hasn't been wiped. An email would get generated, but it'd go to the root account holder, who may not notice in time.
It is unfortunately very rare to find this in practice - everyone seems to be happy with just snapshotting live systems as a backup these days; and it works well enough as long as there is no lingering systemic corruption of data.
(And .... Excel, by mixing data with potentially malicious code, is beyond redemption. But good luck quarantining that in a modern suit controlled company)