Diagnose action results, duplicate writes and enhancement

Understand the likely causes

• Possible causes include trusting hidden IDs, client-only validation, wrong FormData type/cardinality, broad catches swallowing redirect/error control flow, custom enhancement dropping update behavior, or commit before response loss. Kit 3 enhanced status/navigation behavior can differ from old examples. Replays and new edits need different identities even for the same course.

Verify the actual action boundary and protected effect

Identify the exported action and the current principal it receives. Validate hidden IDs, proposed values, revision and request keys at runtime before domain effects. A previous page load or origin check does not prove current object permission. Pair malformed input with a well-formed but unauthorized transition and assert zero protected writer calls for both rejection paths. Keep expected field problems safe and structured through the documented action contract. Do not stringify file-like FormData values into titles or treat repeated fields differently in separate layers. Use synthetic inputs without logging real credentials or full private form bodies.

Compare native and enhanced submission under one version

Confirm method=POST, the targeted page/action convention and whether use:enhance is actually appropriate for that form. A custom callback can change reset, result application and invalidation, so preserve required defaults deliberately. Current Kit 3 behavior for actions targeting another page and fail status differs from older examples; inspect actual status and destination rather than assuming enhanced failures always look like 200. Redirect/error helpers use framework control flow and should not be converted into generic failure by a broad catch. Test expected field feedback and focus with and without JavaScript.

Reconcile a committed result before repeating the effect

The local map records a synthetic operation result and returns it again for the same logical ID without incrementing the effect. It models a desired replay outcome, not durable storage or queue/HTTP delivery. A production protocol must bind identity to principal, action and canonical input, reject changed-input conflicts and handle concurrent duplicates atomically. Define replay retention and crash recovery. Expected revision protects conflicting new edits; operation identity recognizes retransmission. A cache refresh, redirect or client transition can fail after commit, so a generic failure message is not an instruction to repeat the domain write blindly.

const outcomes=new Map();let effects=0;
function attempt(id){if(outcomes.has(id))return outcomes.get(id);const result={revision:++effects};outcomes.set(id,result);return result;}
const first=attempt("rename-17"),replay=attempt("rename-17");
console.assert(first===replay && effects===1);

Repair the earliest faulty contract and retain recovery evidence

Move current authority and validation inside the shared server use case, preserve patch presence and return canonical safe fields or expected problems. Fix enhancement result handling separately from storage replay and source invalidation. Test first success, identical replay, changed input, denied scope, expected conflict and response loss after commit. This guide executes the map only; it does not submit a form, authenticate, invoke Kit or write a source. A real repair needs authorized isolated action/browser/source tests with controlled fakes, safe identities and explicit no-write assertions. Preserve unresolved outcomes rather than inventing rollback from UI failure.

Keep it from recurring

• Use shared domain policy, strict runtime parsing, explicit action-result handling and scoped operation identities. Retain native/enhanced and direct-entry fixtures across Kit migrations. Keep private data out of form results, logs and returned load models, and separate source success from presentation/cache recovery.