Inconsistent Timeout and TimeoutFrame behavior

[TORISOUP]

The behavior of Timeout and TimeoutFrame immediately after calling Subscribe() appears to be inconsistent.
This issue is raised for confirmation in case this behavior is unintended.

TimeoutFrame starts its timer immediately upon subscription. As a result, it times out even if OnNext is never issued.
However, Timeout does not start its timer until OnNext is issued at least once. This means that if OnNext is never issued, the timer will wait indefinitely.

For comparison, UniRx's Timeout starts the timer immediately after subscription.

// R3.Timeout, using NUnit

using var subject = new R3.Subject<int>();

var results =
    subject.Timeout(TimeSpan.FromMilliseconds(100), TimeProvider.System)
        .Materialize()
        .ToLiveList();

// not timeout
await Task.Delay(TimeSpan.FromMilliseconds(300));

// No messages received (subscription is still active)
CollectionAssert.IsEmpty(results);

// R3.TimeoutFrame,  using NUnit

var fakeFrameProvider = new FakeFrameProvider();
using var subject = new R3.Subject<int>();

var results =
    subject.TimeoutFrame(3, fakeFrameProvider)
        .Materialize()
        .ToLiveList();

// Timeout
fakeFrameProvider.Advance(5);

Assert.AreEqual(R3.NotificationKind.OnCompleted, results[0].Kind);
Assert.IsInstanceOf<TimeoutException>(results[0].Error);

// UniRx.Timeout, using NUnit

using var subject = new UniRx.Subject<int>();

var results = new List<UniRx.Notification<int>>();

subject.Timeout(TimeSpan.FromMilliseconds(100))
    .Materialize()
    .Subscribe(results.Add);

// Timeout
await Task.Delay(TimeSpan.FromMilliseconds(200));

Assert.AreEqual(UniRx.NotificationKind.OnError, results[0].Kind);
Assert.IsInstanceOf<TimeoutException>(results[0].Exception);

[neuecc]

Thank you.
In R3, we were thinking of standardizing the timer initialization to occur at the first OnNext across the board.
For example, this applies to Chunk(TimeSpan) and ChunkFrame.
Therefore, I believe the current behavior of TimeoutFrame is a bug, and I'd like to fix it.

[TORISOUP]

I understand that the behavior on the Timeout side is correct and the TimeoutFrame side is currently wrong.
I will keep that in mind for now and use it.