Given a bgzipped VCF that contains a telomere indicated by position 0 (as per the VCF spec):

$ zcat example.vcf.gz | grep -v '#'
22      0       ID1     C       G       .       PASS    .       GT      1/0
22      1       ID2     C       G       .       PASS    .       GT      1/0

(Disclaimer: I know next to nothing about telomeres, and have really only ended up here due to invalid, non-int positions getting converted to 0, as discussed in #1570. My example undoubtedly makes no realistic sense!)

A simple bcftools view works as expected:

$ bcftools view -H  example.vcf.gz
22      0       ID1     C       G       .       PASS    .       GT      1/0
22      1       ID2     C       G       .       PASS    .       GT      1/0

Applying a region filter however unexpectidly drops the telomere (albeit with a warning referring to a flag that I don't think exists in this context?):

$ bcftools view -H --regions 22 example.vcf.gz
[W::tbx_parse1] Coordinate <= 0 detected. Did you forget to use the -0 option?
22      1       ID2     C       G       .       PASS    .       GT      1/0
$ echo $?
0

It seems like the telomere should be included with this filter, given its chr; and I haven't encountered anywhere in the docs that indicates the observed behaviour. However, there certainly could be some biological reason why that doesn't make sense!

The way it gets dropped can cause an interesting issue when considred with #1570, as it means variants where (for whatever reason...) POS gets corruped, get converted to 0 and then dropped when a pipe goes on to have a region filter!