MovieGraphic, There is no video frame in this file.

We have been having an extremely difficult to reproduce bug that's been happening for a few weeks now.
There is a fixation task that we had run with no problems on a computer with near identical hardware dozens of times.
When we switched the task to another computer, without making any changes to the code, it rarely crashes.
The task consists of 4 mov taskobjects, that are looping.
There are only 4 conditions, that just change the order in which the half second videos play.
Here's where things get weird.
The same conditions will be run on repeat for 1000 or so trials, and on most days, it runs with no problems.
However, on some days, halfway through, the program will quit, and this is the error that pops up:

MovieGraphic, There is no video frame in this file.

If the crashes were immediate, or consistent, it would be much easier to diagnose, but its quite strange how this happens.
Do you have any theories for how this could be happening?
The videos are fairly large, as they are uncompressed (to prevent artifacts), at about 100 megabytes each.
I do not think I preloaded the videos in any way, does monkeylogic read the videos into memory from disk, every trial?
(Even though its just the same two videos, in the conditions file)
The newer system has a slightly worse nvme drive (smaller form factor, smaller capacity), would it be possible that gigabytes of reads per minute, over the course of an hour would cause a drive to throttle, thus exiting a session?
I was thinking of upgrading the drive so the two would be identical, for a small cost, but I was wondering if you had a theory of your own.

You should be concerned about the data transfer rate of your disk, since you are using uncompressed videos. I remember that one lab had a similar problem because their disk could not stream video contents at a necessary speed.

NIMH ML does read videos into memory from the disk every trial. You know you use just the same two videos, because you wrote the conditions file, but not all the tasks are written as yours and usually there is no guarantee that the same stimuli are used again.

If that is the case, I would assume software that lets you allocate a portion of your system RAM as a hard drive, would alleviate this problem. (Assuming the total amount of video you are playing per session is less than half your available system RAM)
Or, more easily, if you have 100 dollars available, and a NVME compatible system, making sure you buy a 2280 form factor SSD from a reputable manufacturer, (instead of those square NVME 2240 drives the size of a postage stamp).