#30531 fuzz(restore): diskfull app create probe runs New Workspace as a person

Summary

The diskfull restore fuzz (#30208) gets an app-path create probe. With the disk full, it runs New Workspace the way a person does (debug.toast {perform:"newTab"}, origin user) and checks two things:

• the app keeps no ghost workspace, now and after a relaunch (finding h1);
• the person sees a refusal toast or dialog (finding h3).

Before the probe runs, the harness fills the disk until the daemon refuses a CLI create (margin 0 if it must). If no refusal comes, or the app's create is accepted, the probe is "not run". It is never a pass or a fail in that case. An unknown action id is also "not run".

Harness only: scripts/fuzz/restore/. No product code, and no CI job, step or gate.

Evidence (scheduler rounds, perf minis)

• fgh-h3-head-4203 (#30496 head build): probe pass, one refusal "New Workspace did not apply: the disk is full".
• fgh-h2-head-4202 (build 74bf4924, #30496 head): probe pass; 23 judged, 0 violations.
• fgh-h1-head-4201 (#30268 head, which has no h3 fix): probe fail, "no visible refusal". This is the expected h3 red on a build without the h3 fix.
• STATUS 3393.

Changelog

none

Observed / Not run

• Observed: the three rounds above ran this harness at 855dff1a8d3, which has the same content as this head. prepush-gate: no new violations. py_compile: clean.
• Not run: test_restore_fuzz.py for this head (it is a harness unit test; the rounds ran the code paths).