diff --git a/include/ircd/m/acquire.h b/include/ircd/m/acquire.h
index dcfa29890..0ca811139 100644
--- a/include/ircd/m/acquire.h
+++ b/include/ircd/m/acquire.h
@@ -45,6 +45,9 @@ struct ircd::m::acquire
 	bool fetch_history(event::idx &);
 	void acquire_history();
 
+	bool fetch_timeline(event::idx &);
+	void acquire_timeline();
+
 	bool fetch_state(const m::event::id &, const string_view &);
 	void acquire_state();
 
@@ -74,10 +77,14 @@ struct ircd::m::acquire::opts
 	/// a depth ceiling effectively makes this false.
 	bool head {true};
 
-	/// Perform history acquisition. Setting this to false disables operations
-	/// which fill in gaps in the timeline below the head.
+	/// Perform history acquisition. Setting this to false disables depthwise
+	/// operations which fill in timeline gaps below the head.
 	bool history {true};
 
+	/// Perform timeline acquisition. Setting this to false disables
+	/// breadthwise operations which fill in timeline gaps below the head.
+	bool timeline {true};
+
 	/// Perform state acquisition. Setting this to false may result in an
 	/// acquisition that is missing state events and subject to inconsistency
 	/// from the ABA problem etc.
@@ -109,6 +116,9 @@ struct ircd::m::acquire::opts
 	/// unlimited and that's usually a waste of time in practice.
 	size_t attempt_max {16};
 
+	/// Limit on the depth of leafs pursued by the timeline acquisition.
+	size_t leaf_depth {0};
+
 	/// Default vm::opts to be used during eval; some options are
 	/// unconditionally overriden to perform some evals. Use caution, setting
 	/// options may cause results not expected from this interface.
diff --git a/matrix/acquire.cc b/matrix/acquire.cc
index 0ac9cfee3..c11797a13 100644
--- a/matrix/acquire.cc
+++ b/matrix/acquire.cc
@@ -48,7 +48,7 @@ ircd::m::acquire::acquire::acquire(const struct opts &opts)
 		state_vmopts.wopts.appendix.set(dbs::appendix::ROOM_HEAD, false);
 	}
 
-	if(opts.history)
+	if(opts.history || opts.timeline)
 	{
 		history_vmopts.notify_servers = false;
 		history_vmopts.phase.set(m::vm::phase::NOTIFY, false);
@@ -66,6 +66,10 @@ ircd::m::acquire::acquire::acquire(const struct opts &opts)
 	if(opts.history)
 		acquire_history();
 
+	// Branch to acquire timeline
+	if(opts.timeline)
+		acquire_timeline();
+
 	// Branch to acquire state
 	if(opts.state)
 		acquire_state();
@@ -248,6 +252,176 @@ ircd::m::acquire::fetch_history(event::idx &ref_min)
 	return ret;
 }
 
+void
+ircd::m::acquire::acquire_timeline()
+{
+	event::idx ref_min
+	{
+		opts.ref.first
+	};
+
+	for(size_t i(0); i < opts.rounds; ++i)
+	{
+		if(!fetch_timeline(ref_min))
+			break;
+
+		if(ref_min > opts.ref.second)
+			break;
+	}
+}
+
+bool
+ircd::m::acquire::fetch_timeline(event::idx &ref_min)
+{
+	bool ret(false);
+	auto _ref_min(ref_min);
+	std::set<event::idx> pe;
+	std::deque<event::idx> pq;
+
+	event::idx _event_idx;
+	if(opts.room.event_id)
+		if((_event_idx = m::index(std::nothrow, opts.room.event_id)))
+			pq.emplace_back(_event_idx);
+
+	if(pq.empty())
+		m::room::head(opts.room).for_each([&pq]
+		(const auto &event_idx, const auto &event_id)
+		{
+			pq.emplace_back(event_idx);
+			return pq.size() < event::prev::MAX;
+		});
+
+	if(pq.empty())
+		pq.emplace_back(m::head_idx(opts.room));
+
+	size_t submits(0);
+	size_t leaf_ctr(0);
+	m::event::fetch e, p; do
+	{
+		const auto ref_idx
+		{
+			pq.front()
+		};
+
+		pq.pop_front();
+		if(ref_idx < opts.ref.first || ref_idx < ref_min)
+			continue;
+
+		if(ref_idx > opts.ref.second)
+			continue;
+
+		if(!seek(std::nothrow, e, ref_idx))
+			continue;
+
+		const event::prev prev{e};
+		event::id _prev_id[prev.MAX];
+		const auto &prev_id
+		{
+			prev.ids(_prev_id)
+		};
+
+		assert(prev_id.size() <= prev.MAX);
+		event::idx _prev_idx[prev.MAX];
+		const auto prev_idx
+		{
+			prev.idxs(_prev_idx)
+		};
+
+		size_t fetched(0);
+		for(size_t i(0); i < prev_idx.size(); ++i)
+		{
+			if(prev_idx[i])
+				continue;
+
+			const bool submitted
+			{
+				submit(prev_id[i], opts.hint, false, 1, &history_vmopts)
+			};
+
+			if(!submitted)
+				continue;
+
+			log::debug
+			{
+				log, "Fetch from %s (%lu) miss prev %s fetch:%zu in %s pe:%zu pq:%zu fetching:%zu",
+				string_view{e.event_id},
+				ref_idx,
+				string_view{prev_id[i]},
+				fetched,
+				string_view{opts.room.room_id},
+				pe.size(),
+				pq.size(),
+				fetching.size(),
+			};
+
+			++fetched;
+			++submits;
+			ret |= true;
+		}
+
+		if(pq.size() >= (opts.leaf_depth?: prev.MAX))
+			continue;
+
+		if(opts.leaf_depth || opts.viewport_size)
+		{
+			if(prev_id.size() == 1)
+			{
+				if(++leaf_ctr % (opts.viewport_size?: opts.leaf_depth) == 0)
+					continue;
+			}
+			else leaf_ctr = 0;
+		}
+
+		size_t pushed(0);
+		for(size_t i(0); i < prev_idx.size(); ++i)
+		{
+			if(!prev_idx[i])
+				continue;
+
+			auto it(pe.lower_bound(prev_idx[i]));
+			if(it != end(pe) && *it == prev_idx[i])
+				continue;
+
+			pe.emplace_hint(it, prev_idx[i]);
+			if(!seek(std::nothrow, p, prev_idx[i]))
+				continue;
+
+			pq.emplace_back(prev_idx[i]);
+			_ref_min = std::max(_ref_min, prev_idx[i]);
+			++pushed;
+			log::debug
+			{
+				log, "Queue from %s (%lu) next prev %s (%lu) push:%zu in %s pe:%zu pq:%zu fetching:%zu",
+				string_view{e.event_id},
+				ref_idx,
+				string_view{prev_id[i]},
+				prev_idx[i],
+				pushed,
+				string_view{opts.room.room_id},
+				pe.size(),
+				pq.size(),
+				fetching.size(),
+			};
+		}
+	}
+	while(!pq.empty() && submits < opts.fetch_max);
+
+	log::debug
+	{
+		log, "Round in %s pe:%zu pq:%zu submits:%zu fetching:%zu ref_min:%lu:%lu",
+		string_view{opts.room.room_id},
+		pe.size(),
+		pq.size(),
+		submits,
+		fetching.size(),
+		ref_min,
+		_ref_min,
+	};
+
+	ref_min = std::max(ref_min, _ref_min);
+	return ret;
+}
+
 void
 ircd::m::acquire::acquire_state()
 {