With the addition of the nt dumps using serdi, the weekly run completes on Saturday at around 9:30 pm UTC; that means there's virtually no time when these dumps are not running, no time for recovery if something has to be rerun, no time for maintenance of the host they run on (for example, security updates of kernels plus their reboots), etc. Note that this is before the lexeme dump was added.
What can we do about this?