Unofficial and community-written: I'm not from Supabase, and the tool at the end is my own.
When a Supabase feature "just stops working" and there's no error anywhere, it's often one of the first four below. Each of them returns normally, so nothing in your logs or tests tells you. The fifth does fail with 42501, but only once the call actually runs.
1. A voided query is never sent. supabase-js query builders are lazy, so nothing goes out until something calls .then(). await does that. void doesn't.
void supabase.from('todos').insert({ title }); // never sent
Fix: await it, or chain .then().
2. A subscription to a table that isn't published. If the table isn't in the supabase_realtime publication, the channel reaches SUBSCRIBED and then never receives a postgres_changes event.
alter publication supabase_realtime add table public.messages;
Aman Gupta shares common mistakes developers make when using supabase-js, including issues with lazy query execution, table subscriptions, and row-level security (RLS) policies. He introduces an ESLint plugin to help catch these errors by simulating database migrations and checking client calls. The thread includes additional insights on RLS policies and their impact on operations like UPDATE and INSERT.
Good list. Point 3 is wider than reads, and it trips people up because the fix for one operation depends on a policy for another:
.update(...).eq('id', id) has a WHERE clause, and Postgres applies SELECT policies to the rows it reads. With an UPDATE policy but no SELECT policy, it changes 0 rows with no error..insert(...).select() can fail even when the INSERT policy is fine. The .select() adds RETURNING, and the new row must pass a SELECT policy. If it doesn't, you get 42501 "new row violates row-level security policy". Without .select() the same insert works.A quick way to see silent zero-row updates:
const { data, error } = await supabase
.from('todos').update({ done: true }).eq('id', id).select('id')
if (!error && data.length === 0) console.warn('0 rows matched (RLS or filter)')
And when adding the missing SELECT policy, scope it, e.g. using ((select auth.uid()) = user_id). A quick using (true) exposes every row to every caller.
The behavior is in the CREATE POLICY docs, under "Policies Applied by Command Type": https://www.postgresql.org/docs/current/sql-createpolicy.html
3. RLS on, no select policy. A read comes back as data: [] with error: null. An update or delete changes nothing, again with no error. (Inserts fail loudly with 42501.) The fix is a policy for the role your client uses.
4. .subscribe() with no status callback. CHANNEL_ERROR and TIMED_OUT go nowhere. Pass (status, err) => { ... } and handle them.
5. No grant for the role. From 2026-10-30, new tables in public stop getting automatic grants for anon, authenticated and service_role. A call on a table the role has no grant for fails with 42501, and a Realtime subscription on it connects and receives nothing. Put the grant in the same migration that creates the table. (Tools that replay your migrations only see this change once your project records it: the opt-in migration, or for grants-lint and the plugin below, the since setting in grants-lint.config.json.)
You can spot the first and the fourth from the code alone. The other three need the code and your migrations together, because supabase.from('orders').select() looks fine until you know what the migrations say about orders.
That's why I wrote an ESLint plugin that replays supabase/migrations into a model of the database and checks each client call against it, right in your editor: eslint-plugin-supabase. It doesn't open a network connection or read your .env file. If it can't trace a client, it skips the call. If it can trace the client but can't classify its key (createClient(url, config.key)), it treats it like a publishable-key client by default, and unknownKey: "skip" skips those too. If it flags something that isn't a bug, I'd really like an issue on the repo.
Not affiliated with or endorsed by Supabase.