Possible bug with --resume and .csv --results

[Joe Sapp]

I tried a feature today (version 20210222) that I haven't before and
got an unexpected error:
> $ parallel -j1 --resume --joblog test_joblog.log --results test-$(date 
> +%Y%m%dT%H%M%SZ).csv 'echo {}; sleep 2s' ::: $(seq 100)
> mkdir test-20210303T203326Z.csv: File exists at /home/joe/.local/bin/parallel 
> line 11126.

And the file certainly does (now) exist, but I was not expecting it to
write into it like a directory.

Additionally, this command causes the error:
> $ parallel -j1 --resume --results test-$(date +%Y%m%dT%H%M%SZ).csv 'echo {}; 
> sleep 2s' ::: $(seq 100)

Just '--resume' with --'joblog' works fine (and resumes fine):
> $ parallel -j1 --resume --joblog test_joblog.log 'echo {}; sleep 2s' ::: 
> $(seq 100)
> 1
> 2
> 3

And a plain directory-like argument to '--results' also works fine.

So it looks like it might be the '--resume' with a .csv '--results'
file on a first run.  Am I using or understanding these features
incorrectly?  I would like my script to continue where it left off or
start a new sequence whenever it's run.  Both --joblog and --results
are good features for different purposes.

-- 
Joe