Objection handling

Common pushback and how to answer it. Each answer cites Fly.io’s docs, pricing page, support page, Infra Log or blog, and competitor facts link to the vendor’s own page, all checked 2026-11-04. Where Fly.io itself concedes the point, the answer quotes it and says so. If an answer needs a fact the sources do not settle, it says so.

This page is drawn from https://fly.io/docs/about/pricing/, https://fly.io/docs/about/cost-management/, https://fly.io/docs/postgres/, https://fly.io/infra-log/, https://status.flyio.net/, https://fly.io/blog/corrosion/, https://fly.io/blog/wrong-about-gpu/, https://fly.io/blog/kurt-scott-money-sprites/, https://fly.io/support and the vendor pages named in each answer.

Reliability and company risk

“Fly.io has outages.” True, and Fly.io says so. Its worst outage began on September 1, 2024: “Within a few seconds every proxy in our fleet had locked up hard”, and for a period “no end-user requests could reach our customer apps at all” (https://fly.io/blog/corrosion/). The fix was a watchdog that bounces a deadlocked proxy, and “regionalization” so a state bug threatens a single region (https://fly.io/blog/parking-lot-ffffffffffffffff/, https://fly.io/blog/corrosion/). More recently the Infra Log lists a 15-minute London inbound outage on October 8, 2026, a roughly 75-minute outage for hosted Sprites on October 12, 2026, and about two and a half hours of degraded routing in GRU on September 24, 2026 (https://fly.io/infra-log/). Lead with the honesty: the log is described as “a superset both of our status page events and of customer-impacting events” (https://fly.io/infra-log/).

“What happens when a host fails?” A Machine runs on one physical host and “does not automatically start again on another host” (https://fly.io/docs/blueprints/resilient-apps-multiple-machines/). Run at least two Machines per process; fly launch creates two for processes with services, but only one for apps with volumes (https://fly.io/docs/blueprints/resilient-apps-multiple-machines/). Volumes have no built-in replication between them (https://fly.io/docs/apps/app-availability/).

“Will Fly.io still be around?” Fly.io raised this itself: a well-known reviewer said it was the provider “he pays attention to” he was “least confident would be around by the end of the year” (https://fly.io/blog/kurt-scott-money-sprites/). Fly.io’s answer in the same post is that it raised more money, named a new CEO, and made Sprites “the focus of our company”, while Machines and the platform features are staying (https://fly.io/blog/kurt-scott-money-sprites/). Fundraise size is not stated in the post [verify].

Price

There is no free tier. Correct: “No. New organizations don’t have a free tier or a monthly free usage allowance.” (https://fly.io/docs/about/pricing/). New accounts get a $1 trial credit for 7 days, with limits such as 10 Machines and 5-minute auto-stop (https://fly.io/docs/about/free-trial/). The Sprites site advertises $30 in trial credits, which the docs do not match [verify] (https://sprites.dev/).

“Will I get a surprise bill?” Fly.io concedes the gap: “We don’t support billing alerts (yet), so budget accordingly.” (https://fly.io/docs/about/cost-management/). Budget on the always-on price, which for three shared-1x 1GB Machines in sjc is $23.95/month tops (https://fly.io/docs/about/cost-management/). Customers with any support contract get refunds for unexpected charges “(almost) no questions asked” (https://fly.io/blog/accident-forgiveness/). Dedicated IPv4 at $2/month per app is the usual sneak cost (https://fly.io/docs/about/cost-management/).

Other platforms are cheaper or have free plans. Check the vendor pages. Render lists a Hobby workspace at $0/mo plus compute and a Free instance at $0/month with 512 MB RAM (https://render.com/pricing). Heroku lists Eco at $5, Basic at $7 and Standard-1X at $25 per month, and Eco “Sleeps after 30 minutes of inactivity” (https://www.heroku.com/pricing/). Cloudflare’s Workers Free plan allows 100,000 requests per day (https://developers.cloudflare.com/workers/platform/pricing/). Fly.io’s smallest always-on Machine is $2.19 per 30 days, with no platform fee (https://fly.io/docs/about/pricing/). These are different products, so compare the whole bill: a small web app with Managed Postgres is about $43.67 a month, and the database is most of it (https://fly.io/docs/about/pricing/).

Product scope

You only run containers. Not quite. “Fly.io doesn’t run Docker containers. It uses Docker images as a packaging format”, and the result is a VM booted from an unpacked image (https://fly.io/docs/blueprints/working-with-docker/). Machines can also run multiple containers per VM through the Machines API (https://fly.io/docs/machines/guides-examples/multi-container-machines/). If someone needs a full Linux computer with root, that is Sprites (https://fly.io/blog/design-and-implementation/).

Fly Postgres is not managed. The old one is not: the unmanaged docs say it “is no longer maintained, and Fly.io Support is not able to provide support or guidance for it” (https://fly.io/docs/unmanaged-postgres/). Fly.io’s CEO post lists “unmanaged Postgres (never do this)” among experiments (https://fly.io/blog/kurt-scott-money-sprites/). Managed Postgres (MPG) covers backups, automatic failover, scaling and 24/7 incident response (https://fly.io/docs/postgres/). Be upfront about limits: no in-place major version upgrade, no alerts on CPU or storage, and only some regions (https://fly.io/docs/postgres/, https://fly.io/docs/reference/regions/). MPG has had incidents too, such as the August 1, 2026 IAD etcd overload that blocked new clusters (https://fly.io/infra-log/).

“GPUs were deprecated.” Fly.io wrote “We were wrong about Javascript edge functions, and I think we were wrong about GPUs.” (https://fly.io/blog/wrong-about-gpu/). It said existing GPU Machines were staying, but a v2 should not be expected (https://fly.io/blog/wrong-about-gpu/). As fetched today, https://fly.io/docs/gpus/ and https://fly.io/gpu both redirect to that blog post, and the pricing page lists no GPU line (https://fly.io/docs/about/pricing/). Whether new customers can still provision GPU Machines is unsettled [verify]. Steer to inference APIs, as the post itself recommends (https://fly.io/blog/wrong-about-gpu/).

Scale and limits

“Does it scale?” Known ceilings: a volume maxes at 500GB (https://fly.io/docs/volumes/overview/), MPG at 1 TB of storage (https://fly.io/docs/postgres/), and the Machines API allows 1 request per second per action with bursts to 3 (https://fly.io/docs/machines/api/working-with-machines-api/). Organizations “may be subject to automated scaling limits”, lifted by email (https://fly.io/docs/about/pricing/). A Sprite is fixed at 8 vCPUs and 100 GB (https://fly.io/docs/sprites/concepts/lifecycle/). Fly.io lists 17 regions in the docs (https://fly.io/docs/reference/regions/).

Support

“Is support any good?” The public numbers: Fly.io says it “kept 100% of our response-time promises” and “answered the median ticket in 1h 12m”, and Standard tickets land 55% in 1h and 97% in 24h (https://fly.io/support). But Standard’s promise is a 36-hour first response, Premium 24 hours (1 hour for urgent), Enterprise 4 hours 24x7 (https://fly.io/support). Community support is free (https://fly.io/docs/about/support/). LiteFS and unmanaged Postgres are out of scope (https://fly.io/docs/about/support/).

Lock-in and competition

“Are we locked in?” Mostly no, with caveats. Your app ships as a Docker image (https://fly.io/docs/blueprints/working-with-docker/), and databases move with pg_dump and pg_restore (https://fly.io/docs/postgres/). The sticky parts are fly.toml, volumes, which live on one server in one region, and Fly Proxy features such as fly-replay (https://fly.io/docs/volumes/overview/, https://fly.io/docs/blueprints/multi-region-fly-replay/). Outbound data costs $0.02/GB in North America and Europe (https://fly.io/docs/about/pricing/).

Heroku, Render or Vercel is easier. Fly.io concedes: you can now get “the flexibility of AWS and the ergonomics of Heroku” “here and elsewhere” (https://fly.io/blog/kurt-scott-money-sprites/). Offer a Heroku migration, which Fly.io says takes under 30 minutes for most apps (https://fly.io/docs/getting-started/migrate-from-heroku/). Vercel lists Hobby at $0, Pro at $20/month and a 99.99% SLA at Enterprise (https://vercel.com/pricing); Fly.io’s SLA is 99.9% on Enterprise support (https://fly.io/support).

Agent sandboxes

“Why not a sandbox from Cloudflare, Vercel or Render?” They exist: Cloudflare documents Sandboxes (https://developers.cloudflare.com/sandbox/), Vercel documents Sandbox pricing and quotas (https://vercel.com/docs/sandbox/pricing), and Render announces Sandboxes in beta (https://render.com/pricing). Do not claim a feature gap without reading their docs. Fly.io’s pitch is Sprites: full Linux with checkpoints, an HTTPS URL, Connectors that keep credentials out of the Sprite, and zero compute billing while idle (https://fly.io/sprites/, https://fly.io/docs/sprites/). Caveats: egress is unrestricted by default until you set a network policy (https://fly.io/docs/sprites/concepts/networking/), and Fly.io’s own Infra Log shows Sprites incidents, including an alerting gap it admitted: “We lacked automated alerting for Sprite URLs” (https://fly.io/infra-log/).