Recently, I spoke with Mikita Hanusenka about real-time communications (RTC) to explore the latest developments in the industry and the bottlenecks that still need to be addressed. Mr. Hanusenka is a senior engineering leader with over 10 years of experience who has built and scaled a cross-platform RTC platform for the gaming industry, enabling ultra-low-latency voice and video interactions for millions of concurrent users. He shared his expert perspective on the industry and its key challenges.
Real-time communication (RTC) appears to be a solved problem across the technology industry. Video calls run natively inside browsers without plugins, multiplayer games support tens of millions of concurrent users, and APIs for low-latency streaming have become standardized across platforms. In reality, WebRTC standardized how endpoints establish real-time connections, but it did not standardize how networks behave under congestion, how edge infrastructure responds to global traffic spikes, or how distributed systems maintain synchronization across millions of concurrent participants. As a result, the core challenge of real-time communication has not disappeared. It has evolved.
Mr. Hanusenka adds: ‘The limitations become particularly visible in latency-sensitive environments. A 2021 WebRTC study demonstrated that even relatively small peer meshes struggled to maintain synchronization-grade latency: only 66% of peer connections achieved latency below 30 milliseconds. This threshold is widely considered necessary for high-synchronization applications such as collaborative music or precision interaction. For gaming and other interactive systems, this level of variance can be critical. In competitive first-person shooter (FPS) environments, even 20–40 milliseconds of jitter can disrupt hit registration, movement prediction, and server reconciliation, directly affecting gameplay quality and perceived fairness’.
Major Bottlenecks to Solve
Although the protocol itself has matured, the operational problem has not. WebRTC was originally designed for relatively self-contained applications. However, modern requirements increasingly involve integrating RTC into complex, multi-system interactive environments. This shift forces WebRTC to balance latency, reliability, and resource efficiency. Ensuring synchronization with audio subsystems that are unique for each platform further complicates the situation. The friction particularly rises when attempting to maintain consistent behavior across browsers, mobile devices, and native clients.
Mr. Hanusenka notes that additional practically unresolved bottlenecks include:
- Quality-of-service (QoS) control. It requires continuous adaptation to evolving hardware ecosystems. Performance is tightly coupled to device-specific constraints, meaning systems must be continuously tuned to new chipsets, OSs, and audio pipelines while maintaining at least parity with existing behavior. At the same time, sound quality is inherently subjective and depends on multiple factors, including environmental conditions and sensor sensitivity, further complicating maintenance of consistent performance across diverse hardware.
- Scalability. While establishing real-time ‘rooms’ for small numbers of participants is relatively straightforward, maintaining low-latency synchronization becomes significantly more complex as the number of participants approaches the hundreds and beyond. At that point, the need for multi-layered architecture typically emerges.
Scale Changes the Problem Entirely
The limitations of modern real-time infrastructure become most visible when systems scale from thousands to millions of concurrent users. A clear example that the trend is growing is Roblox, whose total concurrency reportedly exceeded 47 million simultaneous users in August 2025. This surpassed the previous peak recorded by Steam, which reached approximately 41 million concurrent users in March 2025.
‘These numbers matter not only because of traffic volume, but because of the nature of the traffic itself. Real-time gaming systems cannot tolerate delay patterns that are acceptable in traditional web applications. In standard services, additional latency may only slow responses. In real-time systems, however, it can lead to audio disruption or unstable gameplay.
As the number of participants grows, system complexity increases across multiple layers, including replication, matchmaking, event ordering, state synchronization, routing, congestion control, and bandwidth allocation. This produces a ‘fan-out’ effect, where each state update—such as player movement or in-game events—must be distributed to many users across different regions within very short timeframes. Small delays accumulate quickly and degrade overall consistency. These situations often result in ‘thundering herd’ events, where large numbers of clients simultaneously converge on the same infrastructure, creating sharp spikes in load and synchronization pressure,’ admits Mr. Hanusenka.
These failures are rarely caused by full outages. Instead, problems show up as inconsistency. Latency becomes unstable, different regions drift out of sync, voice connections fail intermittently, state updates arrive with delay, performance varies across platforms, and packet loss becomes unpredictable. In practice, the system remains technically available, but the user experience becomes uneven and unreliable across different conditions.
Mikita Hanusenka further explores the problem and notes: ‘Additionally, DDoS protection is becoming increasingly critical for RTC systems. Modern RTC backends typically rely on stateful media servers to minimize latency and often expose a large surface of UDP ports to handle NAT traversal, media routing, and peer connectivity. This creates an inherently larger attack surface compared to traditional HTTP-based systems. Cybercriminals increasingly use this surface using bursty, short-lived attacks to overwhelm infrastructure. For example, in September 2025, Cloudflare stopped a record 11.5 Tbps attack reaching 5.1 billion packets per second, largely dominated by UDP traffic patterns’.
Trade-offs in Real-Time Systems at Scale
Modern RTC systems face conflicting constraints. Lower latency requires tighter, more distributed state sharing, while larger scale pushes systems toward aggregation to reduce complexity. These goals do not align, creating a persistent trade-off between speed and consistency. There is no single design that solves both problems. Improving latency usually increases coordination overhead, while improving scalability adds delay and synchronization drift. As a result, systems do not converge on a fixed solution.
Instead, they adapt dynamically depending on load and network conditions. Under normal usage, they prioritize fast responses. Under stress, they shift toward stability and consistency, even if it increases delay. At scale, real-time systems are less about perfect synchronization and more about managing acceptable inconsistency under changing conditions.
