Exactly. I've never seen a single malware conclusion inside Auto Sandbox and i have never seen Behavior Shield decide that some EXE is malicious using Auto-Decide (default) setting.
I think my concern is perfectly justifiable considering i've seen many behavior blockers perform with outstanding results. Kaspersky PDM, AVG's IDP hell even old and outdated ThreatFire shows more life than avast!'s Behavior Shield ever did (i'm assuming Behavior Shield component is also responsible for Auto Sandbox analysis). Just not avast. Not when it was released with v5, not later with v6 when i expected at least something and yet another letdown during v7 release. Now the Auto Sandbox is taking the part in it and i'm seeing exactly the same pattern.
The feature and idea sounds great but i'm seeing little (or shall i say none) actual results from it.
And now that entire exclusion decision lies on the analysis that aren't even there, i'm concerned, of course.
Before you had to manually exclude, until you did that, the file was always executed inside sandbox.
Now you just slam Continue execution and here we go, executing live malware on a host system.