Should submitting filings be some sort of queue?

[BryceStevenWilley]

Have continually seen Tyler's systems timeout at random times (happened this morning at 6:00am), resulting in the user not being able to submit their filing (they might not be able to easily resubmit from docassemble too, which is a separate issue). It happened twice in the span of about 5 minutes, so that kinda rules out just flakey connections or something similarly transient.

The only way I can think of to handle this issues would be accept all filings in a queue (backed by Postgres, probably), and continually re-try submitting them until Tyler's servers are working again. Would require a very different way of handling the submissions, and necessarily an async way of submitting filings and querying for their status (and more complicated ways of handling their errors (is it an us error, a Tyler error that we know won't resolve, etc.) At least the /status endpoint would be an easy way to extend that behavior (insert a check to the Postgres queue before actually calling Tyler's status SOAP API).

Big idea, not sure if it's worth it, but might be if we really want 100% reliability, instead of 95% (numbers made up).

[nonprofittechy]

This sounds worth implementing to me, given that it isn't easy to submit again from Docassemble.