persist-write-failed

[wizzard] The wizard could not be saved. storage refused the write (<error>). free some space, or
pass a storage of your own; this session will not survive a reload. …/errors/persist-write-failed

A warning from persist() in @wizzard-packages/plugins/persist, printed with console.warn once for the life of the plugin. setItem threw while the plugin was flushing a commit, and <error> carries the error’s name and message.

The usual cause is a full quota (QuotaExceededError): a flow that collects a lot of data, or an origin whose storage other code has already filled. A browser that allows reads and refuses writes produces the same message.

After the first refused write the plugin stops writing for the rest of the session rather than failing on every commit. The wizard is not interrupted: the navigation that committed has already happened, and it stays successful. What is lost is the ability to come back after a reload, which is what the message says.

The same code is printed, with a different cause, when the data itself cannot be written:

[wizzard] The wizard could not be saved. its data cannot be written as JSON (<error>). keep
BigInt values, cycles and class instances out of the data; the last state that could be written
is the one kept. …/errors/persist-write-failed

JSON.stringify threw while the plugin turned a commit into a snapshot - a BigInt, a cycle, or an object whose toJSON throws. That commit is not saved, and nothing else stops: the next commit the plugin can write is saved as usual. The warning is printed once either way, so a storage that refused a write earlier hides this one. The fix is in the data: store a BigInt as a string, and keep anything that is not plain data out of the wizard.

The fix for a refused write is room in the storage, or a storage of your own passed as storage. A flow that holds large values, such as uploaded file contents, is better off storing a reference to them than the values themselves.