# IOTA Train at Home waiting for tasks: how to investigate zero throughput Separate a waiting device from an interrupted data refresh. Check official rosters, run context and history before deciding which local checks to make. - Canonical: https://iotahome.site/en/blog/iota-waiting-for-tasks - Publisher: IOTA Watch - Published: 2026-10-01 - Updated: 2026-10-01 A device can appear in the official miner list while its latest reported throughput is zero. That snapshot is not enough to identify a hardware failure or a lost connection. A useful investigation separates three things: whether the website refreshed, what the official network reported, and what the local training app is doing. ## Check whether the website has current source data Look at the device status source age and any refresh warning first. If the website has not successfully refreshed that source for more than five minutes, IOTA Watch marks the refresh interrupted and retains the prior readings. This indicates a data-read problem; it does not establish that the Mac stopped training. A newer earnings timestamp does not renew a device status timestamp. Different sources can succeed independently. Manual refresh has a cooldown and still respects official rate limits, so repeatedly clicking it is not a reliable way to obtain a new sample. See [refresh interrupted](https://iotahome.site/en/learn/what-refresh-interrupted-means). ## Read the status as a reported sample IOTA Watch uses official miner data to distinguish reported contribution, waiting and non-participation. Throughput belongs to a statistics sample, not a live heartbeat. Check the sample timestamp separately from the time the browser fetched that response. If every active roster was read successfully and an ID is absent, check the full saved Miner ID. If roster coverage is incomplete, an unknown result cannot establish absence. Renaming a device changes its label, not the public ID used for discovery. ## Compare the device with its current run Open the [network page](https://iotahome.site/en/network) and find the run associated with the device. Read roster coverage, slots, reported training counts and progress together. A slot reading does not guarantee immediate assignment, and run-wide progress is not the progress of one Mac. A network snapshot gives context, but it does not explain every local issue. Two devices can be in different runs, and two sources can have different fetch times. Avoid comparing their numbers as if they were synchronized measurements. ## Use history to see whether there is a recent pattern Open device details, select the relevant task and inspect a day or week of throughput and training tokens. Compare several samples rather than one point. History sources are joined by epoch; missing values stay gaps instead of being filled with zero. An empty selected history may mean the API has no records for that task and period. Try a different period, verify the run selection and check source warnings. Neither an empty chart nor a high ranking establishes how much a device should be paid. ## Then make the local checks The [official user guide](https://docs.macrocosmos.ai/product-and-services/tah/tah-user-guide) identifies the app’s readiness status, run panel, run log and update notices as local checks. Open the affected Mac’s app and inspect those views; the monitor cannot read them remotely. - Check the app’s displayed connection and training state - Note the run ID and any visible error or update notice - Compare those observations with the browser’s last successful readings - Use official support when the local app reports an error you cannot resolve Keep the observation precise: “the current sample reports zero throughput” is more useful than assuming the machine is offline. For rewards that look inconsistent with activity, read [zero or unpaid rewards](https://iotahome.site/en/blog/iota-rewards-zero-or-not-paid). ## Sources and implementation - [Macrocosmos · TAH user guide](https://docs.macrocosmos.ai/product-and-services/tah/tah-user-guide) - [IOTA Watch · source code](https://github.com/molimao/iota)