Sometimes /mnt (/dev/sdb1) is 100% full when the build job starts

Describe the bug

In very rare cases, the /mnt disk (/dev/sdb1) is 100% full.
This happens very rarely. Example of such case:
https://github.com/apache/pulsar/actions/runs/16641732202/job/47093641342?pr=24593#step:3:118

  Filesystem     1M-blocks   Used Available Use% Mounted on
  /dev/root         74245M 52119M    22111M  71% /
  tmpfs              7998M     1M     7998M   1% /dev/shm
  tmpfs              3200M     2M     3199M   1% /run
  tmpfs                 5M     0M        5M   0% /run/lock
  /dev/sda15          105M     7M       99M   6% /boot/efi
  /dev/sdb1         75029M 73225M        0M 100% /mnt
  tmpfs              1600M     1M     1600M   1% /run/user/1001

To Reproduce

There isn't a way to reproduce this. It happens rarely. In apache/pulsar actions, I've seen this a few times in the last couple of months.

Expected behavior

/mnt disk should have plenty of available disk space

Runner Version and Platform

Example of where /mnt is 100% full: https://github.com/apache/pulsar/actions/runs/16641732202/job/47093641342?pr=24593#step:3:118

Current runner version: '2.327.1'
Runner Image Provisioner
  Hosted Compute Agent
  Version: 20250711.363
  Commit: 6785254374ce925a23743850c1cb91912ce5c14c
  Build Date: 2025-07-11T20:04:25Z
Operating System
  Ubuntu
  22.04.5
  LTS
Runner Image
  Image: ubuntu-22.04
  Version: 20250728.1.0
  Included Software: https://github.com/actions/runner-images/blob/ubuntu22/20250728.1/images/ubuntu/Ubuntu2204-Readme.md
  Image Release: https://github.com/actions/runner-images/releases/tag/ubuntu22%2F20250728.1

Start with the linked Apache Pulsar job log and compare its runner version, Ubuntu image, and /mnt usage with other affected runs. Trace the hosted runner provisioning context from those details; done means identifying a validated cause for the intermittent full /mnt disk and defining a runner-side correction or mitigation.