There are bugs in the handling of disk cache such that, for instance, at rollover, disk cache stores already built bars of the expiration before the rollover, and then when you do the rollover, it still reloads the old expiration's bars, causing discrepancies versus a machine that did not have cached data before the rollover, or that rolled over at a different time. Additionally, disk cache gets corrupted under various circumstances (many of which seem to have to do with custom bar types), such that the only way to have higher reliability is to clear out the cache before restarting.
Thank you for considering this.

Comment