[Research] A clockless vardiff strands a slowing miner

Testing the resilience of mining hardware and its control systems against various types of disturbances is crucial for ensuring stability and reliability in cryptocurrency mining operations. The testing scenarios range from simple measurement noise to complex grid/power responses, each presenting unique challenges and insights into the system's behavior.

Measurement noise is identified as a primary disturbance, where even stable miners appear jittery due to the Poisson distribution of share submissions. This forms the baseline of disturbances and is crucial for understanding the miner's performance under nominal conditions. Following this, thermal drift occurs as chips heat up and cool down, causing daily performance fluctuations which are important for controllers to monitor and adjust accordingly.

Another significant level of testing involves firmware autotuning, including dynamic voltage and frequency scaling (DVFS) and thermal throttling. Here, miners adjust their hashrate independently of the pool, reacting to thermal changes in real-time. This interaction between miner firmware and external conditions highlights the complexity of maintaining stable operation through automatic adjustments.

Grid and power disturbances, such as demand-response curtailment and real-time pricing throttling, represent deliberate and often scheduled tests. These scenarios typically involve turning machines off and then back on, aligning closely with 100% drop tests but also providing data on recovery processes and system robustness in face of complete power interruptions.

Operational and network events like reboots or hashboard failures mimic hashrate changes at the pool level although they do not affect the actual hashrate. These are critical for testing the observability and responsiveness of the pool's control system, particularly how well it can discern and correct based on the absence or anomalies in data submission.

The core questions revolve around the controller’s ability to correct its behavior based on sparse or absent data, and whether it can accurately reinstate normal operation after disturbances. The concept of decline-safety is introduced, referring to the system's ability to maintain accurate control without error across all possible static states of the miner.

By establishing a clear hierarchy of disturbances and systematically addressing the convergence behaviors at each level, we can better understand the interplay between miner hardware and pool software. This methodical approach ensures thorough testing and validation of the mining system’s resilience to a wide spectrum of realistic operational challenges.