You'll noticeTier: how much it should matter to you
2Useful: my rating, 1 to 5
2Signal: worth watching, 1 to 5
ElsewhereArea: what it touches
ImprovementsKind: in v2.1.286,
ImprovementsSection of the release
What
On a self-hosted runner, meaning a machine you run yourself that carries out remote Claude Code sessions, the part that keeps the inference token fresh can now be asked to refresh straight away. The inference token is the credential the runner uses to make requests to the model. A request to refresh now ends in one of these ways:
a refresh is started
a refresh is already under way
it is skipped because one happened very recently
it is dropped because the session is ending
Each refresh also now reports when its /remote call begins.
Why
This makes it less likely that an expired token holds up a turn on a self-hosted runner.
How sure we are
Something disagreesSomething we can check disagrees with this entry, or the writer said they could not settle it.
The writer flagged doubtWhat triggers an immediate refresh is not shown.
