backlog: file #1294, and record that #1293's recreate-at-path remedy was tested and works

[wshallwshall]

Docs-only: docs/BACKLOG.md.

#1294 -- no cleanup path for the worktrees the prune tool must never touch

Filed explicitly as NOT a coverage bug, because the obvious reading of the measurement is wrong
and would license a dangerous change.

total registered worktrees        36
reachable by prune-merged.ps1     11   (<repo>-<name> siblings)
.claude/worktrees/*               12   excluded BY DESIGN
other (C:/mfw*, Temp, detached)   13   excluded

30 percent reachable. But the exclusions are deliberate and were paid for: that directory is where
EnterWorktree relocates a live session, the tool once removed an occupied worktree and left a
directory git no longer recognised, and its header records the resulting bias as fixed and not
negotiable -- a false SKIP is a minor annoyance, a false PRUNE destroys a session.

The defect is the missing middle. Today a worktree is either removed by the tool or invisible;
nothing produces the list. On 2026-08-19 the owner removed 12 of the excluded ones by hand off a
list assembled manually, which is exactly the work the item proposes to automate -- as a report,
not as a removal. A wider fence was rejected on the fence's own receipt: it cannot see a session
writing by absolute path from elsewhere, 29 percent of writes by primary-seated sessions on this
repo.

The item also states that worktree_gate.ps1's refusal should not be relaxed to close it.

#1293 amendment -- route 1 tested on the live instance

#1293 listed three candidate remedies without picking one. Route 1 was then tested, owner-directed,
against its own live victim: recreate the worktree at the path recorded in the alloc claim, resolve
there, commit there.

The ledger gate reported Passed on a commit that could not have been made from anywhere else,
and PR #397 -- unresolvable by any session ten minutes earlier -- went DIRTY to MERGEABLE.

Bounded in the item: it confirms the mechanism and settles nothing about whether it should be the
documented remedy, since it stays adjacent to the rename-workaround CLAUDE.md section 5 forbids. The
distinction preserved is that it restores the condition the gate tests rather than bypassing the
test.

Two details recorded for anyone repeating it: worktree add -B resets the local branch (here from
91a24061, verified an old main commit reachable from many refs before the reset was accepted),
and the ledger conflict was resolved against a computed target -- base 292/212, branch
293/213, main 310/230, so only 311/231 is correct, which is what parse_items returned. That
is what made "keep both sides" a verified claim rather than a hopeful one.

Ledger control

parse_items: 310/230 to 311/231 for the filing (+1/+1/0), and 311/231 unchanged for the
amendment (0/0/0). Number allocated with alloc.ps1.