In the case of unphased VCF
1/0 are exactly equivalent and the order has no meaning.
Throughout the VCF specification, it is always encouraged to use numerical precedence in order to order the genotypes, e.g.
1/1 would be the preferred representations for diploid bi-allelic genotypes.
However it is not strictly enforced in the VCF specification as for now and you could have
2/0, etc, and bcftools will not complain parsing them. The presence of two representation in the same VCF might indicate that the data comes from multiple variant callers or has been through manual editing, since for example variant callers will always use one representation only (numerical precedence usually)
Also, the order does impact how you read some of the sample information fields, for example the AD (allelic depth), for which the values are read following the order present in the GT field.
I had a look at the VCF 4.2 and 4.3 spec, but indeed is not so precise there on these details.