On 24 September 2026, Cloudflare and researcher Oren Yomtov detailed a vulnerability in Cloudflare Containers and Sandboxes reported on September 4. Yomtov notes that Browser Run used the same disk implementation and was also affected. A Workers Paid customer could recover some residual storage blocks from a previous workload placed on the same host. Fleet remediation, including removal of affected old disks and snapshots, was completed on September 19.
What Cloudflare fixed
The issue involved reused 64 KiB blocks in a shared pool. With skip_block_zeroing enabled, a partial 4 KiB write to a reassigned block could leave the remaining 60 KiB containing data from the previous workload, readable through raw access to the new container’s disk.
Cloudflare says it removed skip_block_zeroing, retired existing container disks and cleared image snapshots created before the fix. It announced that cleanup was complete on September 19. Its review of historical telemetry found only authorized researcher and internal validation activity, with no evidence of malicious exploitation by a third party.
A specific scenario, not a general leak
The scenario required a Workers Paid account and favorable placement. Researchers could not choose a victim, workload or host, and residual data was not guaranteed to be present.
Researchers observed residual data on 18 of 24 production placements they tested, but recovery depended on workload placement and which blocks were reassigned. They could not select a victim, workload or host, or access an actively attached disk. Cloudflare found no evidence that a malicious actor exploited the vector. These limits distinguish possible exposure from a confirmed general breach.
Data to inventory
Cloudflare says customers do not need to make configuration changes. To document sensitive access and organize rotation if another indicator suggests exposure, see our team password-sharing guide and SMB password policy guide. Pay particular attention to secrets, tokens, environment variables and temporary files held in the affected workloads.
Cloudflare’s fix reduces the infrastructure risk. It cannot prove that every secret in a workload remained confidential if other evidence suggests access.
If exposure is possible
If a workload held a sensitive secret and the organization cannot rule out reading, prepare rotation, review logs and preserve useful evidence. Do not replace a key merely to erase a trace; coordinate rotation with the investigation.
Also avoid copying secrets into tickets or incident notes. The remediation record should identify the object and action without reproducing the secret value.
How Soclyde fits
Soclyde does not patch Cloudflare Containers and cannot verify a workload’s history. It can help keep distinct secrets in a local-first encrypted vault and find the ones to rotate after an assessment.
That approach limits centralized copies of access data without turning the vault into proof of infrastructure integrity.
Key takeaway
Cloudflare fixed an isolation flaw that could expose residual blocks, with no malicious exploitation found according to the company. Inventory workload secrets and prepare rotation if needed. See the secure password generator or contact Soclyde.
