We recently encountered a bug in our data usage calculation which has affected the invoices of 25% of our customers for the month of June. As a result of this bug, between June 15 and June 30, Transloadit mistakenly counted every byte towards the usage of affected customers twice.
The bug has now been fixed and invoices for July will be correct. Unfortunately, however, this does mean that we have charged the affected customers too much money for June. We have since gone through all of their Assemblies from that month in order to re-calculate the correct amount the invoices should have specified.
To correct this mistake, we have added the difference as credit to the corresponding accounts. It will be automatically subtracted from the July invoices. We have also emailed all affected customers with details about what usage was calculated and what the actual usage was. We have also offered everybody a refund instead of adding the difference in credit.
The nature of this mistake is very serious, and the consequences are costly. So, in order to prevent this from happening in the future, we have freed our entire core team to scrutinize the responsible code. It turned out that even though the code was very well tested, this was done so by way of simple "unit tests". This meant that in the midst of a big refactoring, individual components would report that they were working well, while in all actuality they were able to produce invalid results when working together.
We have made the necessary improvements to our test suite and added integration tests for this specific situation, so we can now all rest assured that this will never happen again.
We are very sorry that this has happened and offer our sincerest apologies to the affected customers.
If you have any further questions or concerns about this, please don't hesitate to reach out.