When doing a training session today for a group of marine biologists, it became clear that the Sample tool (SPATIAL ANALYST TOOLS> EXTRACTION> SAMPLE) in the latest version of ArcGIS For Desktop (10.4 - downloaded and installed by the participants this week) has a bug in it which means that when you use one raster data layer to extract values from other raster data layers, it populates some data rows with the value -9999. Other values appear to be interpolated with values that are clearly artefacts of the extraction process as they generate impossible values (e.g. negative values for raster cell values of Slope).
This was only a problem with people who were using ArcGIS 10.4 and not those using ArcGIS 10.3 or 10.2. (or indeed 9.3), and occurred on every machine running this version of ArcGIS. It also occurred every time the Sample tool was used with and with multiple different data sets.
While there is a work around by turning the raster data layer into a point data layer and then using the Extract Multi Values To Points tool (SPATIAL ANALYST TOOLS> EXTRACTION> EXTRACT MULTI VALUE TO POINTS), this issue means that the standard protocols used to analyse marine biological data when producing species distribution models (SDMs) no longer function as they should. Worse than this, this problem doesn't return an error message, but rather generates artefacts within the resulting data set without any warning meaning it is all too easy for people to assume that the data set created by the tool is reliable when it isn't, and since SDMs are commonly used to make important management and conservation decisions, this has serious real-world implications.
My question is this: Is there anywhere that such bugs can be reported so they can be fixed ASAP? At the moment, there is a lot of pressure to shift across to open source alternatives to ArcGIS (such as QGIS), and suddenly encountering situations where previously good tools in ArcGIS no longer function as they should and instead produce data sets with serious, but hard to detect, artefacts, means that most of those in the training session were left feeling they would be better off using QGIS rather than ArcGIS because the appropriate tools in QGIS worked as they should without generating such artefacts (and without having to use awkward, non-standard, work-arounds). It doesn't help that there appears to be no standard protocol for reporting such bugs, and so no way of providing re-assurance as to when it it might be sorted.
Any feedback on what do to about such situations would be gratefully appreciated as I was left at a loss as to what to tell the people I was meant to be training.