In `@crates/biome_js_analyze/src/lint/nursery/use_member_ordering.rs`:
- Around line 90-96: The diagnostic currently emits a hardcoded note in the
.note(markup! { "Members should be ordered by: properties, constructors,
methods." }) call which is misleading for non-default or user-configured orders;
update the code in use_member_ordering.rs to either remove this static note or
build the note text dynamically from the active ordering settings (the rule's
configured order) and use that generated string in the .note(...) invocation
(refer to the place where .note(markup! { ... }) is called and the state/options
that contain the ordering to compose the message).
- Line 54: The rule registration currently uses
RuleSource::EslintTypeScript("member-ordering").same() which incorrectly claims
identical behavior; change it to
RuleSource::EslintTypeScript("member-ordering").inspired() to indicate we only
mirror some class-member ordering behavior (not full
interface/type/enum/memberTypes/optionalityOrder/order support). Update the
entry where sources: &[RuleSource::EslintTypeScript("member-ordering").same()]
is defined to use .inspired() so the rule source accurately reflects
partial/modified behavior.
- Around line 427-451: In find_order_violations, prev_rank currently stores the
last seen member's rank which lets lower-ranked violating members lower the
baseline and mask later violations; change the logic so prev_rank tracks the
maximum rank seen so far (and its label) instead of the immediate predecessor:
when you compute rank = group.rank(order), if rank < prev_r push the
MemberOrderViolation (using member_range and expected_before_name), but do NOT
set prev_rank to the lower rank—instead set prev_rank = Some((max(prev_r, rank),
label_for_max)) so prev_rank always holds the highest rank and its corresponding
label; update references to prev_label/prev_r accordingly in the loop.

In `@crates/biome_js_analyze/src/lint/nursery/use_member_ordering.rs`:
- Around line 88-100: The current deduplication uses
label.split_whitespace().last() which truncates meaningful group names (symbols:
seen, order, label, broad, order_description) — change this to use the full
group label (or map each MemberGroup variant to an explicit, human-friendly
broad category) and dedupe while preserving order: iterate over order, derive
the full label or mapped category from the MemberGroup variant instead of using
split_whitespace().last(), track seen categories with a HashSet (or Vec +
contains) to avoid duplicates, push the full category strings into seen, and
then join seen into order_description so the resulting message uses clear,
complete group names.
- Around line 455-467: The current logic uses group.rank(order) and treats the
sentinel value order.len() like a normal rank, which causes unlisted groups to
inflate max_rank and wrongly trigger MemberOrderViolation; change the comparison
in the member-order loop so that when group.rank(order) == order.len() you skip
that member (do not compare against or assign to max_rank and do not push a
violation for it), leaving max_rank unchanged for subsequent listed groups;
reference the symbols max_rank, group.rank(order), MemberOrderViolation,
member_range and expected_before_name when locating and updating the code.

In `@crates/biome_js_analyze/src/lint/nursery/use_member_ordering.rs`:
- Around line 210-219: The TS getter/setter signature branches hardcode
is_hash_private=false; instead determine hash-private from the member name like
other cases: call is_hash_private_name on the signature's name (e.g. use
getter.name() / setter.name() to get the AnyJsClassMemberName and pass it to
is_hash_private_name) and pass that boolean into resolve_get_accessor_group /
resolve_set_accessor_group; keep existing is_static
(has_static_modifier_ts_method) and accessibility logic intact.

In `@crates/biome_js_analyze/src/lint/nursery/use_member_ordering.rs`:
- Around line 152-210: classify_member() currently only returns GetAccessor and
SetAccessor but the MemberGroup enum and default_order() include combined
variants like Accessor/StaticAccessor/ProtectedAccessor that will never be
produced; either remove those combined Accessor variants and their ranks from
MemberGroup and default_order() or, if they are reserved for TC39 auto-accessor
support, add a clear comment on the enum and in default_order() explaining they
are placeholders for the proposal and why classify_member() doesn't produce
them; update any code that matches on these combined variants (e.g.,
default_order(), ranking logic) to reflect the removal or to document planned
future support so custom orders like ["property","accessor","method"] behave as
expected.
- Around line 58-88: The rustdoc lists member group names in camelCase (e.g.,
indexSignature, hashPrivateProperty, staticBlock) but the actual deserialised
names are defined by the MemberGroup enum via #[serde(rename = "...")] and the
Deserializable macro (which expects kebab-case and `#private-`… forms); update the
documentation comment above use_member_ordering.rs to list the exact JSON keys
that the enum accepts (the serde-renamed variants such as index-signature,
static-index-signature, property, static-property, `#private-property` or
`#private-`… notation used in the enum, constructor, static-block, etc.) so copied
examples match the MemberGroup enum/Deserializable expectations (refer to
MemberGroup enum and the Deserializable macro for the canonical names to
document).