Methods and systems for reliable token revocation at a server are described. The server receives, a token revocation policy, which includes an identification of a set of users for which a set of associated tokens are to be revoked. The server receives, from a first client device, a first request to access resources at the server, the first request including a first token generated at the token authority server for the client device, wherein the first token is associated with a first expiration time interval; and denies access to the resources at the server based on the first token and the token revocation policy, prior to an expiration of the first expiration time interval associated with the first token.