I've limited the use of the correlation ID to the teacher advisor sign-up endpoint (as discussed) to limit the potential for issues if any problems occur. I've also not been able to best the filter attribute, it is possible but I cannot mock the hangfire context because they've neglected to apply interfaces, or make class methods generic. We could use a library like Typemock but not sure if this will play nicely with the CICD pipeline out of the box.
If theses changes prove stable we can raise a second, follow-on ticket to complete the work to apply this logging behaviour across the remaining endpoints.