Talk to all your customers.

Ask your customers who are leaving. Don't guess their reasons for leaving. Actually ask all of them directly.

Ask the new customers who are signing up, why they are signing up?

Ask the customers who are not leaving, what they use it for, what they like about the service and what they don't. Corroborate with your own metrics.

It's a temporary phase. You'll get over this.

Are you still feeling burnt out? Is this a project you want to continue on? Selling the business is not a bad thing if it's not where you want to continue putting your time.

Are cancellations outnumbering new registrations because cancellations are increasing (relative to total customers), or because signups are decreasing? If the former, asking customers why they cancel is certainly something you should do. I'd also take a look at your analytics, and look for usage patterns you can draw insights from. Install mixpanel, and set a 'canceled' super-property on any user that cancels. Then you can retroactively segment their usage vs ongoingly active users.

If you want to let the service go on auto-pilot, you need customer acquisition methods that enable it. AdWords, paying someone to write / promote blog posts, a referral program, etc.

I think you already have your answer.

> Coincidentally, I have been working less (hacking away at the inessential) by being way more selective in implementing features, experimenting way too much with pricing, and toning down my marketing (which basically consists on participating on forums and publishing blog articles)

Working 100 hours a week isn't sustainable, but neither is working an hour a day when your baby is still trying to find its footing. Experimenting a lot with pricing has its benefits as an educational tool but it's also an expensive one which may result in what's happening to your subscription numbers. Then again you won't really know if that was the real answer unless you survey your customers.

Your other submitted questions may answer this for you.. for instance: http://news.ycombinator.com/item?id=3455292

' I have no growth strategy to be honest, and I just feel like living a Tim Ferriss 4-hour-work-week for the entire year ahead.'

Attempt to make contact with every single person that cancelled. Catalog their concerns, there will be patterns. This that aren't glaringly obvious to you but are visible to the customer. Address those concerns with existing customers to stem the flow. Adjust. Rinse. Repeat.

Review where your previous influx of customers came from then put a little effort back in.

Maybe you should be also looking after your existing customers e.g gathering feedback, thanking them, offering awesome support instead of tinkering with the core of the product.

link?