What's the speed bottleneck here? Are you performing thousands of queries of this type? Or are you performing a smaller number of very large queries?

If you're performing lots of queries, but the number of variants and samples is not that large, bcftools query should be fine, as long as your BCF is indexed.

If you're performing a smaller number of very large queries, and you only have biallelic variants, plink 1.9 is likely to outperform bcftools. You can convert your data to plink-format with

plink --bcf <full BCF filename> --const-fid 0 --out converted ;

your sample query can then be executed with

plink --bfile converted --keep-allele-order --chr 1 --from-bp 1 --to-bp 15 --recode rlist --out result .

See https://www.cog-genomics.org/plink/1.9/formats#rlist for a description of the format of this command's main output file (result.rlist). Positional information for the variants can be looked up in converted.bim, or if you use a plink 1.9 build from 30 Nov 2019 or later, result.map (--recode rlist forgot to generate the .map file in older plink 1.9 builds).

using bioalcidaejdk: http://lindenb.github.io/jvarkit/BioAlcidaeJdk.html

 java -jar dist/bioalcidaejdk.jar -e 'stream().forEach(V->println(V.getContig()+ " "+V.getStart()+ " "+V.getReference().getDisplayString()+" "+V.getAlternateAlleles().stream().map(A->A.getDisplayString()).collect(Collectors.joining(","))+" "+V.getGenotypes().stream().filter(G->!(G.isNoCall() || G.isHomRef())).map(G->G.getSampleName()).collect(Collectors.joining("\t"))));' src/test/resources/rotavirus_rf.vcf.gz


RF01 970 A C S5
RF02 251 A T S2 S3
RF02 578 G A S4
RF02 877 T A S1
RF02 1726 T G S2    S3
RF02 1962 TACA TA S1
RF03 1221 C G S2    S3
RF03 1242 C A S4
RF03 1688 T G S5
RF03 1708 G T S5
(...)