Self-hosting downsides are real and specific: patching, email deliverability, and backups nobody has actually tested are the three things that break. What does not break is quieter, and it is the half of this argument most posts skip: nobody can sunset your board out from under you, and your bill does not grow because you succeeded.
What's the strongest argument against self-hosting?
Not a hypothetical one. On 29 October 2025, someone on Second Life's public feedback board filed a request titled "Replace Canny with Selfhosted Alternative", proposing LogChimp on cost and data-ownership grounds. Maestro Linden declined it officially.
The comments are the part worth reading in full, because they carry the honest counter-case rather than a dismissal. Linden Lab would become entirely responsible for patching and stability if the board moved in-house, and the team had already moved to Canny.io from self-hosting once before, specifically to get out from under that burden. That is not a theoretical trade-off raised by a vendor's marketing page. It is a team that tried both directions and named, in public, the reason they went the way they went.
Take it seriously rather than arguing around it: for a team with nobody who wants to own patching and stability, self-hosting genuinely is the worse choice, and Linden Lab's own history is real evidence of that, not a strawman built to be knocked down.
What genuinely breaks when you self-host?
Security patching
Fider, the best-documented open-source option in this category, shipped three security releases inside three months: v0.35.0 fixed a critical authentication vulnerability, v0.36.0 fixed SSRF and cross-site scripting, and v0.36.1 carried further security work. Each of those has to be noticed and applied by someone. A hosted vendor absorbs the same list of releases without a customer ever seeing a version number. If you would rather not be the one watching for the next one, Fider alternatives covers what else is available and how each handles this. The same applies to anything wired up around the board: an integration that quietly stops posting looks exactly like a quiet week until somebody checks.
Email deliverability
Fider's own setup requirements list your own SMTP provider or Amazon SES as something you supply. Getting a sending domain configured so a mail provider will actually trust it, SPF, DKIM and DMARC records set up correctly, is real, specific work, and it fails silently: the first sign is usually a customer mentioning their reply notification landed in spam. SMTP for self-hosted tools walks through getting this right the first time.
Backups nobody has restored
The most common failure mode in this category is not dramatic. A backup exists, and nobody has ever tried to restore it. A database dump that has never been test-restored is not a backup, it is an assumption. This is where the cheapest form of self-hosting, one small VPS running everything, is most exposed: if the box holding your data is also the only place a restore could run from, a bad afternoon can turn into a bad month.
One payment, no subscription, unlimited products.
What does not break?
This is the half of the story most self-hosting posts leave out.
Nobody can sunset your board out from under you. Feedback tools in this category have genuinely closed: Olvy announced its closure in January 2025 and shut on 15 February 2025. Pendo's classic Feedback product, formerly Receptive.io, was sunset on 1 August 2026. Delighted blocked renewals from July 2025 and closed in June 2026. GetFeedback Direct is scheduled for discontinuation on 31 December 2026. Self-hosted code does not carry a sunset date set by someone else's roadmap. An abandoned project stops receiving patches, which is a real cost covered above, but the software already running keeps running.
Your bill does not grow because your board succeeded. Canny.io's tracked-user count and Featurebase's per-seat charge both rise with the exact thing you were hoping for: more people using the board. A self-hosted instance costs the same on the day it has ten voters as the day it has ten thousand.
Your export is not gated behind a tier you didn't buy. Canny.io's own CSV export carries post data without comments; comments need a separate JSON API call. Since 12 September 2025 the EU Data Act has capped the switch away from a data processing service at 30 calendar days, but that clock only starts after a notice period the contract may set at up to two months, and no vendor in this category mentions any of it. We are not lawyers and this is not legal advice, and whether a feature request board counts as a data processing service under the Act is a judgement we are not qualified to make. On a self-hosted instance, none of it arises, because you already hold the database directly.
If there is no server, most of this list does not apply at all. Docket, for one example, runs as a static site plus two small endpoints, so there is no Postgres to patch, no container to keep alive and no SMTP server to misconfigure, and the posts are files in a repository that is also the backup. That is an architectural choice, not a universal fix for self-hosting: it only removes the risk for tools built that way, and most self-hosted options in this category, Fider, LogChimp and Quackback among them, are not.
Read the two lists against each other before deciding either way. What breaks is real. What does not break is just as real, and it is why self-hosting stays a genuine option rather than a mistake waiting to happen. For the full cost picture behind both lists, see self-hosting a feature request board.
Frequently asked questions
What was the strongest real case made against self-hosting?
Second Life's public feedback board, 29 October 2025: a request to replace Canny.io with a self-hosted LogChimp instance, declined officially by Maestro Linden, with comments noting Linden Lab had already moved to Canny.io from self-hosting once before, specifically to reduce the patching and stability burden.
What is the single most common self-hosting mistake?
Never test-restoring a backup. A database dump nobody has restored is an assumption, not a backup, and it is usually only discovered during the incident it was meant to prevent.
Does self-hosting mean nobody can shut my board down?
Nobody can sunset the software itself the way a vendor can. Several hosted tools in this category have genuinely closed, including Olvy in February 2025 and Pendo's classic Feedback product in August 2026. A self-hosted instance keeps running even if the project behind it stops being maintained, though it also stops receiving security patches at that point.
How often do security patches actually happen in this category?
Fider, the most active open-source option, shipped three separate security releases inside three months: a critical authentication fix, an SSRF and cross-site scripting fix, and a further security release. That is the realistic cadence to plan around, not a one-off.
Do all self-hosted boards need a server to run?
Most do. Fider, LogChimp and Quackback are conventional web applications needing a database and a host. A board built as a static site does not, because the posts are files and the pages are generated at build time, which removes the patching, deliverability and backup risks above rather than managing them.