My own take: you can't trust any web-based solution. But you can choose a provider which plays nice with open standards, supports eff and openpgp etc.
I don't let mailfence handle my keys, but I liked that they on every level, let me do what I want.
And I like that they let other people trust them to handle their pgp keys. Thus helping that ecosystem.
I guess Electron might be a reasonable application to quickly move your crypto code from the website to a local client so that a compromised server can't simply backdoor it. But then you want it to auto-update for security issues and you're back to square one. And if someone inspects the source code of each update (they won't), they'll just be slower updating and either run the old (vulnerable) code for longer, have forced and unscheduled downtime (while looking at the diff), or run the new code before it has been vetted.
Also, their "we take software security" blurb is weak:
We use operating systems and open source software that take security seriously. However, software have bugs. In most cases, an update for a security problem will be available within minutes/hours of the original report. We perform the update as soon as it is available and validated.
Applying patches is table stakes. What portions of their stack, including their own code, have they actually had audited? Do they have software security engineers on staff? Is there a /security page somewhere on this site that explains where to report vulnerabilities?