I happened to be readying a new reference genome and was using FastaStats to force the creation of the .dict and .fai files. The automatic .dict file creation first creates a temporary file, creates the .dict using Picard, and then copies it as appropriate. However, the newer versions of Picard won't overwrite the dict file, and so there is an error using the Java created temporary file as an output. The problematic section seems to be ReferenceDataSource.java. The error manifests as:
Index file /gs01/projects/ngs/resources/gatk/2.3/human_g1k_v37.dict does not exist but could not be created because: . /gs01/projects/ngs/resources/gatk/2.3/dict8545428789729910550.tmp already exists. Delete this file and try again, or specify a different output file.