Pings from instances to nodequery are 98% on the low end and 99.9% on the high end. It looks like I'm at 500+ days up at impactvps.com, though I used to see cpu and disk degradations somewhat regularly with them. I haven't had any major outages in a few years of use, though one provider went bust recently after an overly-aggressive sale (alpharacks).

I generally run one instance + sqlite and accept the downtime. The next step up might be two vms across two providers + a cheap managed db.

I don't think I have any super strong recommendations. I did become a fan of BuyVM recently due to their $1/month shared sql plan. But, they're apparently they're planning to retire it, which puts them at a similar price point as something like vultr.