HAL known issues list 
As of: September 27, 2012

 * FileProblemInstances (i.e. problem instances defined by a file on the filesystem) will only ever be defined once, with their initial path on the filesystem. Subsequent instance definitions using a file with the same hash, stored elsewhere on the filesystem, will use the original definition and path. This does not cause problems in HAL, but can cause user confusion in the filesystem structure when defining several instance lists containing the same instance, and then exporting and re-importing in another filesystem.

 * When creating a configuration experiment, once a user has made the decision to clone an existing experiment or define a new one, that choice cannot be reverted without refreshing the page.

 * When creating an algorithm, once a user has made the decision to clone an existing algorithm or define a new one, that choice cannot be reverted without refreshing the page.

 * When creating a new configuration, once a user has made the decision to create a new configuration for an existing parameterized algorithm or to clone the configuration used in a parameterless algorithm, that choice cannot be reverted without refreshing the page.

 * The experiment subruns and HAL log browsing pages have an "auto-refresh" functionality that is very confusing in the UI currently. We know, and are working to change it soon.

* In our Torque cluster, Orcinus, we are currently noticing that the recording of output data from target algorithm runs to the database is not being done until after the meta algorithm is finished.  This does not affect the results of the experiment, but statistics and plots will not be available until the experiment is finished.

* When creating images for plots, each plot has a unique filename.  Leaving browsers on webpages which generates plots (all status pages. ie. analysis and configuration statuses) can quickly fill the server's file system with old image data.  Restarting the server will remove all this old data.