Come up with a strategy for ensuring robustness/reliability of blobs
We have focused a lot in soledad/blobs performance latelly, and not so heavily in robustness. "Slow is ok as long as it doesn't loose documents", so we have to bring the focus back to reliability/robustness and come up with items that we can check.
Some notes:
- we currently have a spec and ongoing implementation for crash recovery: #8926 (closed)
- how to increase our guarantee of robustness? Some ideas follow:
- backups
- defending against multi-device issues
- a simple app to import/export/view the local database (so it's not a black box to the user)
- we should start prioritizing robustness over performance to make sure we are on the right track.