>>So you're leaving a W7 VM running on VirtualBox continuously for weeks at a time? Neither W7 nor VirtualBox were really designed for that.
>
>It's pretty much the same period of time as when I ran W7 on its own box. It would crash differently, though.
>
>>There was a period of time in the 3.x days when no further work was being done on VirtualBox, it was a bit moribund. At that time the impression I got from various articles was that it was a decent desktop/workstation hypervisor but not really suitable for heavy production use. Since then they've made good progress with the 4.x and 5.x branches but you still never hear anyone running production server loads on it the way they're run on other hypervisors.
>
>I'm pushing my SQL hard. I'm doing some conversions (almost finished now!) and it's loading megabytes at a time, then reloading from backup, fixing a few things and running again. But if SQL gets slow, just restart the service (takes only a few seconds) and it's fast again.
>
>But I guess the combined memory thrashing that it gets from so much SQL and the screen savior (which also has a little database of its own, as your tagline says or used to say, dbfs only though) will get it on its knees after a while.

If you're working with tens of thousands of images I wonder if you're actually hitting a GDI resource limit: https://blogs.msdn.microsoft.com/dsui_team/2013/04/23/debugging-a-gdi-resource-leak/

ISTR there was some command that needed to be run when putting image files in reports on the fly, so that the correct image would actually appear for each detail band rather than just the first image loaded being cached and repeated endlessly. Ah - remembered it - CLEAR RESOURCES. If it turns out to be a GDI resource issue you might want to issue that command before/after loading each new image.