Yes pouchdb does not get the love it deserves. But it is still actively maintained, the last bug I had was timely fixed by the community and merged.

I feel the other way around; pouch is pretty rock solid at this point and I've not had trouble throwing data into it or around between pouch installs. Meanwhile every single couchdb server and installation has vastly different limits from each other, from obvious after the fact limits (but would have been nice to know much earlier in architecture planning) such as very limited attachment sizes and very poor attachment upload/download/sync performance, to much less obvious limits that seem incredibly arbitrary and underdocumented such as pouch and localhost Apache Couch 1/2 are very flexible in terms of database and document naming (supporting a large character set and mostly ignoring things), and Couchbase, Cloudant, and Apache Couch 3 all have very different allowed character sets and are differently strict about certain names.

On top of that hosting any of the server applications is increasingly a pain to setup/manage especially from a PaaS approach rather than an IaaS approach. IBM has only done terrible things to make this worse over time. My requirements when I started was I needed to host on Azure, and Cloudant supported that at the time. IBM of course being IBM and focusing entirely on IBM Cloud and its new name/strategy/plan every ~nine months dropped the Azure support I'm still told I need. But the constant name changes (IBM BlueMix, IBM Cloud, whatever it will be next week, IBM WatsonRain or whatever), plan changes, cheese moving, don't give me a lot of confidence in IBM's Cloud efforts even if I wasn't feeling a lot of pressure from my IT colleagues to get everything (back) into Azure. I'm almost desperate enough to build a Pouch driver for CosmosDB myself at this point.