[Jim Young <jyoung@xxxxxxx>]

> On Wednesday, December 2, 2015 14:09 <guy@xxxxxxxxxxxx>
> On Dec 2, 2015, at 8:11 AM, Evan Huus <eapache@xxxxxxxxx> wrote:
> > My current hypothesis is commit 74541a9596eead6647c592de9aa46797c2dffa84 but I don't have any files to test with locally.
>
> So that one looks as if it might affect *startup* time but not *file loading* time.

I have a trace on a Windows 8 system that appears to replicate the issue.   The file loads in ~8 sec for release 2.0.0 and ~25.5 sec on master.  The initial load time with when the trace has apparently been flushed from the file I/O system cache adds an additional ~10 sec:  ~18.5sec on 2.0.0 vs ~34.5 on master.

I'm bisecting now.