diff --git a/src/mcp/mcp.c b/src/mcp/mcp.c
index 7016a0d2..ea595ce8 100644
--- a/src/mcp/mcp.c
+++ b/src/mcp/mcp.c
@@ -789,20 +789,42 @@ static cbm_store_t *resolve_store(cbm_mcp_server_t *srv, const char *project) {
     project_db_path(project, path, sizeof(path));
     srv->store = cbm_store_open_path_query(path);
     if (srv->store) {
-        /* Check DB integrity — auto-clean corrupt databases */
+        /* Check DB integrity — preserve corrupt databases instead of
+         * silently deleting them so the user can inspect the evidence and
+         * recover or report the issue (#557). Rename to .corrupt.<ts> so
+         * the path is obvious in the cache directory. */
         if (!cbm_store_check_integrity(srv->store)) {
-            cbm_log_error("store.auto_clean", "project", project, "path", path, "action",
-                          "deleting corrupt db — re-index required");
             cbm_store_close(srv->store);
             srv->store = NULL;
-            /* Delete the corrupt DB + WAL/SHM files */
-            cbm_unlink(path);
+            /* Rename the corrupt DB + WAL/SHM to preserve evidence */
             char wal_path[MCP_FIELD_SIZE];
             char shm_path[MCP_FIELD_SIZE];
+            char corrupt_path[MCP_FIELD_SIZE];
+            char corrupt_wal[MCP_FIELD_SIZE];
+            char corrupt_shm[MCP_FIELD_SIZE];
+            long long now_s = (long long)time(NULL);
             snprintf(wal_path, sizeof(wal_path), "%s-wal", path);
             snprintf(shm_path, sizeof(shm_path), "%s-shm", path);
-            cbm_unlink(wal_path);
-            cbm_unlink(shm_path);
+            snprintf(corrupt_path, sizeof(corrupt_path), "%s.corrupt.%lld", path, now_s);
+            snprintf(corrupt_wal, sizeof(corrupt_wal), "%s-wal.corrupt.%lld", path, now_s);
+            snprintf(corrupt_shm, sizeof(corrupt_shm), "%s-shm.corrupt.%lld", path, now_s);
+            if (rename(path, corrupt_path) != 0) {
+                cbm_unlink(path); /* fallback: delete if rename fails */
+            }
+            /* best-effort: sidecars may not exist */
+            if (rename(wal_path, corrupt_wal) != 0) {
+                cbm_unlink(wal_path);
+            }
+            if (rename(shm_path, corrupt_shm) != 0) {
+                cbm_unlink(shm_path);
+            }
+            cbm_log_error("store.auto_clean", "project", project, "path", path, "action",
+                          "corrupt db preserved", "preserved_as", corrupt_path);
+            (void)fprintf(stderr,
+                          "ERROR corrupt project database: %s\n"
+                          "ERROR   preserved as: %s\n"
+                          "ERROR   inspect the .corrupt file, delete it when done, and re-index.\n",
+                          path, corrupt_path);
             return NULL;
         }
 
diff --git a/src/pipeline/pipeline.c b/src/pipeline/pipeline.c
index 8e370f7c..24d39477 100644
--- a/src/pipeline/pipeline.c
+++ b/src/pipeline/pipeline.c
@@ -783,13 +783,19 @@ static int try_incremental_or_delete_db(cbm_pipeline_t *p, cbm_file_info_t *file
     if (!db_path) {
         return CBM_NOT_FOUND;
     }
-    struct stat db_st;
-    if (stat(db_path, &db_st) != 0) {
+    /* Open the existing DB directly in query mode (never creates the file).
+     * A NULL result means it is absent or unreadable — nothing to reindex
+     * from, so we return without touching the path. Deriving existence from
+     * the open instead of a prior stat() removes the check-then-use race on
+     * db_path that CodeQL flags as a TOCTOU, and matches resolve_store() in
+     * mcp.c, which already uses this pattern. */
+    bool was_corrupt = false;
+    cbm_store_t *check_store = cbm_store_open_path_query(db_path);
+    if (!check_store) {
         free(db_path);
         return CBM_NOT_FOUND;
     }
-    cbm_store_t *check_store = cbm_store_open_path(db_path);
-    if (check_store && cbm_store_check_integrity(check_store)) {
+    if (cbm_store_check_integrity(check_store)) {
         cbm_file_hash_t *hashes = NULL;
         int hash_count = 0;
         cbm_store_get_file_hashes(check_store, p->project_name, &hashes, &hash_count);
@@ -806,17 +812,45 @@ static int try_incremental_or_delete_db(cbm_pipeline_t *p, cbm_file_info_t *file
             cbm_log_info("pipeline.route", "path", "mode_change_reindex", "stored_hashes",
                          itoa_buf(hash_count), "discovered", itoa_buf(file_count));
         }
-    } else if (check_store) {
+    } else {
+        /* Integrity check failed — this is the data-loss path of #557. */
+        was_corrupt = true;
         cbm_store_close(check_store);
     }
-    cbm_log_info("pipeline.route", "path", "reindex", "action", "deleting old db");
-    cbm_unlink(db_path);
-    char wal[PL_WAL_BUF];
-    char shm[PL_WAL_BUF];
-    snprintf(wal, sizeof(wal), "%s-wal", db_path);
-    snprintf(shm, sizeof(shm), "%s-shm", db_path);
-    cbm_unlink(wal);
-    cbm_unlink(shm);
+    char wal_path[PL_WAL_BUF];
+    char shm_path[PL_WAL_BUF];
+    snprintf(wal_path, sizeof(wal_path), "%s-wal", db_path);
+    snprintf(shm_path, sizeof(shm_path), "%s-shm", db_path);
+    if (was_corrupt) {
+        /* Preserve the corrupt DB as .corrupt.<timestamp> instead of deleting,
+         * so the user can inspect/recover before the full reindex (#557).
+         * Only on real corruption — a healthy DB rebuilt due to a mode change
+         * is expected churn and must not accumulate .corrupt files. */
+        long long now_s = (long long)time(NULL);
+        char corrupt_db[PL_WAL_BUF];
+        char corrupt_wal[PL_WAL_BUF];
+        char corrupt_shm[PL_WAL_BUF];
+        snprintf(corrupt_db, sizeof(corrupt_db), "%s.corrupt.%lld", db_path, now_s);
+        snprintf(corrupt_wal, sizeof(corrupt_wal), "%s-wal.corrupt.%lld", db_path, now_s);
+        snprintf(corrupt_shm, sizeof(corrupt_shm), "%s-shm.corrupt.%lld", db_path, now_s);
+        if (rename(db_path, corrupt_db) != 0) {
+            cbm_unlink(db_path);
+        }
+        if (rename(wal_path, corrupt_wal) != 0) {
+            cbm_unlink(wal_path);
+        }
+        if (rename(shm_path, corrupt_shm) != 0) {
+            cbm_unlink(shm_path);
+        }
+        cbm_log_info("pipeline.route", "path", "reindex", "action", "corrupt db preserved",
+                     "preserved_as", corrupt_db);
+    } else {
+        /* Healthy DB being rebuilt (mode change / file-count drift) — delete. */
+        cbm_log_info("pipeline.route", "path", "reindex", "action", "deleting old db");
+        cbm_unlink(db_path);
+        cbm_unlink(wal_path);
+        cbm_unlink(shm_path);
+    }
     free(db_path);
     return CBM_NOT_FOUND;
 }
diff --git a/tests/test_mcp.c b/tests/test_mcp.c
index 68edc0e9..3d4a0740 100644
--- a/tests/test_mcp.c
+++ b/tests/test_mcp.c
@@ -9,6 +9,9 @@
 #include <mcp/mcp.h>
 #include <store/store.h>
 #include <yyjson/yyjson.h>
+#include <sqlite3.h> /* sqlite3_exec — corrupt-DB regression #557 */
+#include <dirent.h>  /* opendir/readdir — corrupt-DB regression #557 */
+#include <sys/stat.h>
 #include <string.h>
 #include <stdlib.h>
 #include <stdbool.h>
@@ -2209,6 +2212,86 @@ TEST(tool_bad_project_name_no_overflow_issue235) {
 }
 #undef ISSUE235_DBNAME
 
+/* Issue #557: a corrupt project DB must be PRESERVED (renamed to
+ * <name>.db.corrupt.<ts>), not silently unlink()'d. resolve_store()
+ * runs the integrity check on first query; a numeric root_path trips
+ * it (same trigger as store_integrity_corrupt_bad_path). After the
+ * query, the original .db must be gone but a .corrupt sidecar must
+ * exist so the user can inspect/recover. */
+TEST(resolve_store_preserves_corrupt_db_issue557) {
+    char cache[256];
+    snprintf(cache, sizeof(cache), "/tmp/cbm-557-XXXXXX");
+    if (!cbm_mkdtemp(cache)) {
+        PASS(); /* skip if mkdtemp fails */
+    }
+
+    const char *saved = getenv("CBM_CACHE_DIR");
+    char *saved_copy = saved ? strdup(saved) : NULL;
+    cbm_setenv("CBM_CACHE_DIR", cache, 1);
+
+    /* Build a structurally-valid DB whose projects row is corrupt
+     * (numeric root_path), so cbm_store_check_integrity() returns false. */
+    char db_path[512];
+    snprintf(db_path, sizeof(db_path), "%s/regr557.db", cache);
+    cbm_store_t *seed = cbm_store_open_path(db_path);
+    ASSERT_NOT_NULL(seed);
+    sqlite3_exec(cbm_store_get_db(seed),
+                 "INSERT INTO projects (name, indexed_at, root_path) "
+                 "VALUES ('regr557', '2024-01-01', '826');",
+                 NULL, NULL, NULL);
+    cbm_store_close(seed);
+
+    /* A query against the project triggers resolve_store → integrity check. */
+    cbm_mcp_server_t *srv = cbm_mcp_server_new(NULL);
+    ASSERT_NOT_NULL(srv);
+    char *resp = cbm_mcp_server_handle(
+        srv, "{\"jsonrpc\":\"2.0\",\"id\":1,\"method\":\"tools/call\",\"params\":{\"name\":"
+             "\"search_graph\",\"arguments\":{\"label\":\"Function\","
+             "\"project\":\"regr557\"}}}");
+    ASSERT_NOT_NULL(resp);
+    free(resp);
+    cbm_mcp_server_free(srv);
+
+    /* The original DB must be gone (renamed, not left in place). */
+    struct stat st;
+    ASSERT_TRUE(stat(db_path, &st) != 0);
+
+    /* A <name>.db.corrupt.<ts> sidecar must have been created. */
+    bool found_corrupt = false;
+    DIR *d = opendir(cache);
+    ASSERT_NOT_NULL(d);
+    struct dirent *ent;
+    while ((ent = readdir(d)) != NULL) {
+        if (strncmp(ent->d_name, "regr557.db.corrupt.", 19) == 0) {
+            found_corrupt = true;
+        }
+    }
+    closedir(d);
+    ASSERT_TRUE(found_corrupt);
+
+    if (saved_copy) {
+        cbm_setenv("CBM_CACHE_DIR", saved_copy, 1);
+        free(saved_copy);
+    } else {
+        cbm_unsetenv("CBM_CACHE_DIR");
+    }
+    /* Clean up any regr557.db* files left in the temp cache dir. */
+    DIR *cd = opendir(cache);
+    if (cd) {
+        struct dirent *e;
+        while ((e = readdir(cd)) != NULL) {
+            if (strncmp(e->d_name, "regr557.db", 10) == 0) {
+                char p[768];
+                snprintf(p, sizeof(p), "%s/%s", cache, e->d_name);
+                cbm_unlink(p);
+            }
+        }
+        closedir(cd);
+    }
+    cbm_rmdir(cache);
+    PASS();
+}
+
 /* ══════════════════════════════════════════════════════════════════
  *  SUITE
  * ══════════════════════════════════════════════════════════════════ */
@@ -2353,4 +2436,5 @@ SUITE(mcp) {
     RUN_TEST(snippet_include_neighbors_enabled);
     RUN_TEST(snippet_source_invalid_utf8);
     RUN_TEST(tool_bad_project_name_no_overflow_issue235);
+    RUN_TEST(resolve_store_preserves_corrupt_db_issue557);
 }