Currently the InformationLayer.GetBestView method takes in account visible size of the markers (in pixels) and makes best possible assumption about its geographical size when calculating location rectangle. But size in pixels corresponds to different geographical size (in latitude/longitude degrees) depends on the zoom level. So it could take 2-3 times to get really optimal zoom level depends on the initial zoom level. The InformationLayer.GetBestView method should be changed to provide precise result at once regardless of the initial zoom level.