no-task-tracker-references reports every hit no-process-history-references already refuses, and since 1.17.0 it reads every file, so consumers disable it

[HackingGate]

What consumers do with the rule

Across a fleet of about 90 policy files, 14 list no-task-tracker-references
in inherit.disabled_rules and 3 more shadow it with a same-id copy that
restates the pattern to narrow files.exclude. The comments give three
recurring reasons. The first is duplication, from a workspace policy:

# Disabled because no-process-history-references, inherited from the
# same set, already refuses every linking form this one does ... and
# running both reports each hit twice.

The second is the 1.17.0 widening. Two consumers under one superproject:

# uphold 1.17.0 widened no-task-tracker-references from Markdown, reST
# and plain text to every file in the tree ... the first run under the
# wider glob reported 27 of them on a pointer bump that touched none.

and a sibling counted 276 across src/, packaging and scripts on a commit
that touched none of them. Both kept the rule as a docs-only copy.

The third is a superproject whose source comments cite an issue as the
place a measurement lives; it disables both rules in 9 policy files with
the same sentence: "An issue reference in this tree is an evidence anchor
rather than tracker state leaking into prose."

Why

policy/base/process-residue.toml line 119 onward: no-task-tracker-references
has files.include = ["."] and no files.glob since commit 0f79c26 (#158,
shipped in 1.17.0), which removed files.glob = ["*.md", "*.rst", "*.txt"].
no-process-history-references at line 134 still carries that glob. The
two overlap on the github.com/<owner>/<repo>/(issues|pull)/N URL and on
issue #N, which the tracker rule's prose-word arm also matches, so a
Markdown line in either form is reported twice under two ids. The
workspace comment overstates the overlap (the history rule does not match a
bare #N), but the double report on the shared forms is real. PR #192 adds a separator arm to the bare-number alternation, which widens the
rule again over every source file the 1.17.0 glob change admitted.

What would close it

Split by scope: keep no-task-tracker-references on the docs glob it had
before 1.17.0 and ship the every-file scope as a second id a consumer opts
into, or the reverse with the docs id as the opt-in. And de-duplicate
against no-process-history-references so a line both patterns match is
reported once, naming both ids. Each of the 14 disables was a consumer
choosing between a rule that refuses every commit and no rule at all.