I contacted UCSC and was forwarded to the NCBI Genome Reference Consortium, since the gap files are based on AGP files from GRC. This is the official response in case anyone is curious:
Dr. Page and his lab have long served as collaborators of the GRC. As
a consequence, we were made aware of the location of the mouse
centromere way back in 2009, and did include it in GRCm38. As
described in the Soh et al publication, the centromere is a ~90 kb
region found in the BAC clone AC175459.4 (bp 57238-147035). That
clone is included in the tiling path for the GRCm38 assembly, and the
corresponding position of the centromere in the chromosome is
If you examine the assembly AGPs (the files which describe the FASTA;
https://ftp.ncbi.nlm.nih.gov/genomes/all/GCA/000/001/635/GCA_000001635.8_GRCm38.p6/GCA_000001635.8_GRCm38.p6_assembly_structure/Primary_Assembly/assembled_chromosomes/AGP/), you'll see that the 3 Mb gaps at the start of each chromosome have
distinct specifications. On all chromosomes except Y, the first 100
kb represent the telomere, the next 10 kb represent the short arm, and
the next 2.89 Mb, the centromere. On chr. Y, the gap is only 101000
bp, and represents only the telomere and short arm.
The chromosome Y centromere is not listed in the AGP files because
this centromere is not a gap- it is sequenced. The AGP format only has
mark-up available for biological gaps (telomere, centromere, short
arm, heterochromatin), and does not support a similar mark-up if those
regions have been sequenced. For GRCm39, we will be using a separate
file to explicitly define the chr Y centromere.
The first sequenced base (non-N) of chr Y does not include telomeric
repeats, so we know that we’re still missing some sequence at this
end. Thus, we still include a gap to account for this “missing”
sequence, and we have used the default short-arm length for this for
consistency with the other chromosomes.
Although it does not fully answer my question, it substantially clarifies the issue.