$ bedops --merge --range 2000 <(gff2bed < foo.gff) > merged.bed
$ bedmap --count --echo-map --delim '\t' merged.bed <(gff2bed < foo.gff) | awk '($1>0)' | cut -f2- | sort-bed - | bedops -n 1 <(gff2bed < foo.gff) - > unmerged.bed

Or, in one pass, using standard IO streams:

$ bedops --merge --range 2000 <(gff2bed < foo.gff) | bedmap --count --echo-map --delim '\t' - <(gff2bed < foo.gff) | awk '($1>0)' | cut -f2- | sort-bed - | bedops -n 1 <(gff2bed < foo.gff) - > unmerged.bed

Might be faster to convert once:

$ gff2bed < foo.gff > foo.bed

Then do the ops:

$ bedops --merge --range 2000 foo.bed | bedmap --count --echo-map --delim '\t' - foo.bed | awk '($1>0)' | cut -f2- | sort-bed - | bedops -n 1 foo.bed - > unmerged.bed

Thank you, Alex. I tried your syntax, first converting GFF to BED, simply using gff2bed. And then,

bedops --merge foo.bed | bedmap --count --echo-map - foo.bed | awk '($1>0)' | cut -f2- | sort-bed - | bedops -n 1 foo.bed - > unmerged.bed

I received an error message that reads:

End coordinate is too large. Max decimal digits allowed is 12 in BEDOPS.Constants.hpp. See line 1 in -.

Perhaps my BED input file is non-standard? For example, see below:

bedops --merge foo.bed | bedmap --count --echo-map - foo.bed | awk '($1>0)' | cut -f2- | head -n 1

837036 877324 ORF68_tool_Prfl1 8 - tool_Prfl1 ORF 2 ID=ORF68_ptool_Prfl1

Error message looks like it is generate at the sort-bed using STDIN stream step...

bedops --merge foo.bed | bedmap --count --echo-map - foo.bed | awk '($1>0)' | cut -f2- | sort-bed -

So I've tried reverting back to bedtools merge and it looks like it does report BOTH merged and unmerged coords in the same output file. I am now making sure I am not mis-interpreting mergeBed with bleary eyes, using examples and real data!