diff --git a/src/BLite.Core/Storage/LockTimeout.cs b/src/BLite.Core/Storage/LockTimeout.cs
index 10693990..a2959821 100644
--- a/src/BLite.Core/Storage/LockTimeout.cs
+++ b/src/BLite.Core/Storage/LockTimeout.cs
@@ -60,15 +60,23 @@ namespace BLite.Core.Storage;
     public static LockTimeout Immediate => new() { ReadTimeoutMs = 0, WriteTimeoutMs = 0, MaxConcurrentWriters = 0 };
 
     /// <summary>
-    /// Balanced default: 500 ms write timeout, 500 ms read timeout,
-    /// 96 max concurrent writers.
+    /// Balanced default: 5 s write timeout, 5 s read timeout, 96 max concurrent writers.
     /// <para>
-    /// 500 ms is long enough for commits under normal load (observed p99 well under 500 ms)
-    /// but short enough to fail-fast when the engine is saturated rather than leaving
-    /// threads blocked for seconds, starving the thread-pool.
+    /// A writer waits for other writers' I/O, not only for their CPU work: an auto-committing
+    /// write holds the collection lock (in single-file mode, typed collections share one lock for
+    /// the whole file) until its commit is in the WAL, and a checkpoint holds the page-file write
+    /// lock while it syncs the data file. Writers queued behind each other wait for several such
+    /// writes, and a disk can stall for hundreds of milliseconds. 5 s matches the WAL's own lock
+    /// timeout, so a <see cref="TimeoutException"/> means a writer that is stuck (a deadlock or a
+    /// hung disk), not one that is slow.
+    /// </para>
+    /// <para>
+    /// The asynchronous write paths wait with <see cref="System.Threading.SemaphoreSlim.WaitAsync(int)"/>,
+    /// which blocks no thread while waiting. Use <see cref="Immediate"/> for fail-fast
+    /// (<c>SQLITE_BUSY</c>) behaviour, or <see cref="From"/> for other values.
     /// </para>
     /// </summary>
-    public static LockTimeout Default => new() { ReadTimeoutMs = 500, WriteTimeoutMs = 500, MaxConcurrentWriters = 96 };
+    public static LockTimeout Default => new() { ReadTimeoutMs = 5_000, WriteTimeoutMs = 5_000, MaxConcurrentWriters = 96 };
 
     /// <summary>
     /// Creates a <see cref="LockTimeout"/> from explicit millisecond values.
diff --git a/src/BLite.Core/Storage/PageFile.cs b/src/BLite.Core/Storage/PageFile.cs
index 36b42b63..dc259ef9 100644
--- a/src/BLite.Core/Storage/PageFile.cs
+++ b/src/BLite.Core/Storage/PageFile.cs
@@ -46,7 +46,8 @@ public readonly struct PageFileConfig
     /// Lock acquisition timeouts for read and write operations.
     /// Controls how long the engine waits for a contended lock before throwing
     /// <see cref="System.TimeoutException"/> (analogous to SQLite's <c>SQLITE_BUSY</c>).
-    /// Defaults to <see cref="LockTimeout.Default"/> (500 ms read / 500 ms write).
+    /// The <see cref="Default"/>, <see cref="Small"/>, <see cref="Large"/> and <see cref="Server"/>
+    /// presets use <see cref="LockTimeout.Default"/> (5 s read / 5 s write).
     /// Use <see cref="LockTimeout.Immediate"/> for fail-fast behaviour.
     /// </summary>
     public LockTimeout LockTimeout { get; init; }
diff --git a/src/BLite.Core/Storage/StorageEngine.cs b/src/BLite.Core/Storage/StorageEngine.cs
index 4c5d256b..82c3cd8d 100644
--- a/src/BLite.Core/Storage/StorageEngine.cs
+++ b/src/BLite.Core/Storage/StorageEngine.cs
@@ -483,7 +483,7 @@ public void Dispose()
         // its truncation and frees fail, leaking the pages for good. Wait for it to finish (#209).
         //
         // The wait lasts as long as that checkpoint, not a lock timeout: on a loaded CI runner one
-        // took longer than WriteTimeoutMs (500 ms) and was cut short while releasing the pages. It
+        // took longer than WriteTimeoutMs (then 500 ms) and was cut short while releasing the pages. It
         // cannot hang: every lock the checkpoint takes (commit lock, cross-process writer lock,
         // page file locks) is waited for with its own timeout, its I/O is finite (the WAL images,
         // one flush, the truncation, one free per queued page), it waits for nothing Dispose
diff --git a/tests/BLite.Tests/IndexPageReleaseTests.cs b/tests/BLite.Tests/IndexPageReleaseTests.cs
index df8d2353..52688c70 100644
--- a/tests/BLite.Tests/IndexPageReleaseTests.cs
+++ b/tests/BLite.Tests/IndexPageReleaseTests.cs
@@ -334,9 +334,8 @@ public async Task Dispose_DuringPendingFreeCheckpoint_DoesNotLeakTheQueuedPages(
     [Fact]
     public async Task Dispose_WaitsForThePendingFreeCheckpoint_LongerThanTheLockTimeout()
     {
-        var storage = new StorageEngine(_dbPath, PageFileConfig.Default);
+        var storage = new StorageEngine(_dbPath, PageFileConfig.Default with { LockTimeout = LockTimeout.From(500, 500) });
         var timeout = storage.LockTimeout.WriteTimeoutMs;
-        Assert.InRange(timeout, 1, 1000);
         var checkpoint = new TaskCompletionSource<bool>(TaskCreationOptions.RunContinuationsAsynchronously);
         typeof(StorageEngine).GetField("_pendingFreeCheckpointTask", Flags)!.SetValue(storage, checkpoint.Task);
 
diff --git a/tests/BLite.Tests/LockTimeoutDefaultTests.cs b/tests/BLite.Tests/LockTimeoutDefaultTests.cs
new file mode 100644
index 00000000..7065ad5d
--- /dev/null
+++ b/tests/BLite.Tests/LockTimeoutDefaultTests.cs
@@ -0,0 +1,179 @@
+using System.Collections.Concurrent;
+using System.Diagnostics;
+using System.Reflection;
+using BLite.Core;
+using BLite.Core.Storage;
+using BLite.Core.Transactions;
+using BLite.Shared;
+
+namespace BLite.Tests;
+
+/// <summary>
+/// #303: a writer waits for the commits queued ahead of it, and each commit waits for the disk, so
+/// <see cref="LockTimeout.Default"/> must outlast several slow commits, and a timeout set on
+/// <see cref="PageFileConfig.LockTimeout"/> must reach every lock that waits with it.
+/// </summary>
+public class LockTimeoutDefaultTests : IDisposable
+{
+    private const BindingFlags Flags = BindingFlags.NonPublic | BindingFlags.Instance;
+    private readonly string _dir = Path.Combine(Path.GetTempPath(), $"locktimeout_{Guid.NewGuid():N}");
+
+    public LockTimeoutDefaultTests() => Directory.CreateDirectory(_dir);
+
+    public void Dispose()
+    {
+        try { Directory.Delete(_dir, recursive: true); } catch { /* best-effort */ }
+    }
+
+    [Fact]
+    public void Default_IsFiveSecondsForReadsAndWrites()
+    {
+        Assert.Equal(LockTimeout.From(readTimeoutMs: 5_000, writeTimeoutMs: 5_000, maxConcurrentWriters: 96), LockTimeout.Default);
+    }
+
+    [Fact]
+    public void Presets_AndDetectedConfigs_UseTheDefault()
+    {
+        Assert.Equal(LockTimeout.Default, PageFileConfig.Small.LockTimeout);
+        Assert.Equal(LockTimeout.Default, PageFileConfig.Default.LockTimeout);
+        Assert.Equal(LockTimeout.Default, PageFileConfig.Large.LockTimeout);
+        Assert.Equal(LockTimeout.Default, PageFileConfig.Server(Path.Combine(_dir, "server.db")).LockTimeout);
+
+        foreach (var (name, config) in new[] { ("small", PageFileConfig.Small), ("large", PageFileConfig.Large) })
+        {
+            var path = Path.Combine(_dir, name + ".db");
+            new BLiteEngine(path, config).Dispose();
+            Assert.Equal(LockTimeout.Default, PageFileConfig.DetectFromFile(path)!.Value.LockTimeout);
+        }
+
+        using var memory = BLiteEngine.CreateInMemory();
+        Assert.Equal(LockTimeout.Default, memory.Storage.LockTimeout);
+    }
+
+    /// <summary>
+    /// A configured timeout reaches the WAL, the main, index and collection page files, the
+    /// admission gate and both kinds of collection.
+    /// </summary>
+    [Fact]
+    public async Task ConfiguredTimeout_ReachesEveryLock()
+    {
+        var timeout = LockTimeout.From(readTimeoutMs: 1_234, writeTimeoutMs: 2_345, maxConcurrentWriters: 7);
+        var path = Path.Combine(_dir, "engine", "test.db");
+        Directory.CreateDirectory(Path.GetDirectoryName(path)!);
+        using (var engine = new BLiteEngine(path, PageFileConfig.Server(path) with { LockTimeout = timeout }))
+        {
+            var col = engine.GetOrCreateCollection("docs");
+            await col.InsertAsync(col.CreateDocument(["value"], b => b.AddInt32("value", 1)));
+            var storage = engine.Storage;
+
+            Assert.Equal(timeout, storage.LockTimeout);
+            Assert.Equal(2_345, (int)Field(Field(storage, "_wal"), "_writeTimeoutMs"));
+            Assert.Equal(timeout, PageFileLockTimeout(Field(storage, "_pageFile")));
+            Assert.Equal(timeout, PageFileLockTimeout(Field(storage, "_indexFile")));
+            var collectionFiles = (ConcurrentDictionary<string, Lazy<IPageStorage>>)Field(storage, "_collectionFiles");
+            Assert.NotEmpty(collectionFiles);
+            foreach (var file in collectionFiles.Values)
+                Assert.Equal(timeout, PageFileLockTimeout(file.Value));
+            Assert.Equal(7, ((SemaphoreSlim)Field(storage, "_writerGate")).CurrentCount);
+            Assert.Equal(2_345, WriteLockTimeoutMs(col));
+        }
+
+        var ctxPath = Path.Combine(_dir, "context", "test.db");
+        Directory.CreateDirectory(Path.GetDirectoryName(ctxPath)!);
+        using (var db = new TestDbContext(ctxPath, PageFileConfig.Default with { LockTimeout = timeout }))
+            Assert.Equal(2_345, WriteLockTimeoutMs(db.TestDocuments));
+    }
+
+    /// <summary>
+    /// The commit lock waits for the configured write timeout, not for the default: with the lock
+    /// held, an auto-committing insert fails after about 150 ms, well before the 5 s default.
+    /// </summary>
+    [Fact]
+    public async Task CommitLock_WaitsForTheConfiguredWriteTimeout()
+    {
+        using var engine = new BLiteEngine(Path.Combine(_dir, "commit.db"),
+            PageFileConfig.Default with { LockTimeout = LockTimeout.From(150, 150) });
+        var col = engine.GetOrCreateCollection("docs");
+        await col.InsertAsync(col.CreateDocument(["value"], b => b.AddInt32("value", 0)));
+
+        var commitLock = (SemaphoreSlim)Field(engine.Storage, "_commitLock");
+        await commitLock.WaitAsync();
+        try
+        {
+            var sw = Stopwatch.StartNew();
+            var ex = await Assert.ThrowsAsync<TimeoutException>(async () =>
+                await col.InsertAsync(col.CreateDocument(["value"], b => b.AddInt32("value", 1))));
+            sw.Stop();
+            Assert.Contains("commit lock", ex.Message);
+            Assert.InRange(sw.ElapsedMilliseconds, 100, 3_000);
+        }
+        finally
+        {
+            commitLock.Release();
+        }
+    }
+
+    /// <summary>
+    /// Writers on one collection queue behind each other's commits, and each commit here takes
+    /// 150 ms to reach the WAL (a disk sync on a slow disk). The last of 8 writers waits about a
+    /// second, twice the 500 ms the default used to allow, and none of them may time out.
+    /// </summary>
+    [Fact]
+    public async Task WritersQueuedBehindSlowCommits_DoNotTimeOut()
+    {
+        const int writers = 8;
+        const int commitMs = 150;
+        var wal = new SlowFlushWriteAheadLog(new MemoryWriteAheadLog(), commitMs);
+        var pages = new MemoryPageStorage(16384);
+        pages.Open();
+        using var engine = BLiteEngine.CreateFromStorage(new StorageEngine(pages, wal));
+        var col = engine.GetOrCreateCollection("docs");
+        await col.InsertAsync(col.CreateDocument(["value"], b => b.AddInt32("value", -1)));
+
+        wal.Slow = true;
+        var start = new TaskCompletionSource<bool>(TaskCreationOptions.RunContinuationsAsynchronously);
+        var inserts = Enumerable.Range(0, writers).Select(w => Task.Run(async () =>
+        {
+            await start.Task;
+            var sw = Stopwatch.StartNew();
+            await col.InsertAsync(col.CreateDocument(["value"], b => b.AddInt32("value", w)));
+            return sw.ElapsedMilliseconds;
+        })).ToList();
+        start.SetResult(true);
+        var elapsed = await Task.WhenAll(inserts);
+
+        Assert.True(elapsed.Max() > 500, $"the slowest writer took {elapsed.Max()} ms; the scenario must wait longer than the old 500 ms default");
+        Assert.Equal(writers + 1, (await col.FindAllAsync().ToListAsync()).Count);
+    }
+
+    private static object Field(object owner, string name)
+        => owner.GetType().GetField(name, Flags)!.GetValue(owner)!;
+
+    private static LockTimeout PageFileLockTimeout(object pageFile)
+        => ((PageFileConfig)Field(Assert.IsType<PageFile>(pageFile), "_config")).LockTimeout;
+
+    private static int WriteLockTimeoutMs(object collection)
+        => (int)collection.GetType().GetProperty("WriteLockTimeoutMs", Flags)!.GetValue(collection)!;
+
+    /// <summary>Delays every flush, like a WAL whose commit waits for a slow disk sync.</summary>
+    private sealed class SlowFlushWriteAheadLog(IWriteAheadLog inner, int delayMs) : IWriteAheadLog
+    {
+        public volatile bool Slow;
+
+        public async Task FlushAsync(CancellationToken ct = default)
+        {
+            if (Slow) await Task.Delay(delayMs, ct).ConfigureAwait(false);
+            await inner.FlushAsync(ct).ConfigureAwait(false);
+        }
+
+        public ValueTask WriteBeginRecordAsync(ulong transactionId, CancellationToken ct = default) => inner.WriteBeginRecordAsync(transactionId, ct);
+        public ValueTask WriteCommitRecordAsync(ulong transactionId, CancellationToken ct = default) => inner.WriteCommitRecordAsync(transactionId, ct);
+        public ValueTask WriteAbortRecordAsync(ulong transactionId, CancellationToken ct = default) => inner.WriteAbortRecordAsync(transactionId, ct);
+        public ValueTask WriteDataRecordAsync(ulong transactionId, uint pageId, ReadOnlyMemory<byte> afterImage, CancellationToken ct = default)
+            => inner.WriteDataRecordAsync(transactionId, pageId, afterImage, ct);
+        public long GetCurrentSize() => inner.GetCurrentSize();
+        public Task TruncateAsync(CancellationToken ct = default) => inner.TruncateAsync(ct);
+        public List<WalRecord> ReadAll() => inner.ReadAll();
+        public void Dispose() => inner.Dispose();
+    }
+}