Ian Cooper says event-driven orgs of up to about five teams can manage async APIs on shared knowledge of each other's services. His model for doing it at scale begins with the address, binding and contract each endpoint has to describe to its consumers.
Reality
- Evidence62
- Adoption
- Insufficient
- Hype gap+10
- Incentives20
- Confidence55
Monest's CTO kept tag-event fan-out on BullMQ, where each job goes to one worker, for two client integrations and several internal teams. A first attempt, one worker calling every consumer, replayed finished calls on retry and forced one retry count on all of them.
Reality
- Evidence40
- Adoption8
- Hype gap+8
- Incentives55
- Confidence45
A dev.to post argues that moving embeddings out of Postgres buys a second source of truth, a four-leg query path and 50 to 150 ms of internet latency. Its 8 ms pgvector counter-figure has no benchmark behind it.
Reality
- Evidence24
- Adoption
- Insufficient
- Hype gap+55
- Incentives55
- Confidence45
A dev.to walkthrough puts the hard part of background job retries in recovery semantics. The test it sets is whether a restarted worker can tell finished work from work it merely fetched. The queue's own limits are narrow.
Reality
- Evidence24
- Adoption
- Insufficient
- Hype gap+28
- Incentives76
- Confidence46
The queue layer publishes fast and drains slowly, so a store can confirm every order while inventory_reservation quietly grows. A tuning guide on dev.to points the audit at the broker and the consumer roster instead of the frontend.
Reality
- Evidence52
- Adoption
- Insufficient
- Hype gap+12
- Incentives55
- Confidence55
RabbitMQ 4.2.2 drops global QoS and classic queue mirroring, so a Celery task with an ETA parks its worker until the ETA arrives. Because a queue's type is fixed at declaration, the fix has to land in the application first.
Reality
- Evidence45
- Adoption30
- Hype gap+25
- Incentives25
- Confidence50
Collapsing notification, scheduling and retry into one Postgres job table cut roughly 4,100 lines to about 600 without losing a capability, and it deleted a dual write that could drop a receipt in silence.
Reality
- Evidence45
- Adoption15
- Hype gap+20
- Incentives35
- Confidence50
Hardcoded algorithms and forced TLS 1.2 across 90-plus repositories stop OpenStack inheriting post-quantum crypto the platform already ships. Keystone's token signing is one hardcoded line.
Reality
- Evidence58
- Adoption32
- Hype gap+18
- Incentives72
- Confidence52
Putting the job queue in the customer database costs no new infrastructure and hands autovacuum a race it loses. The benchmark cited against it caps Postgres near 660 messages a second.
Reality
- Evidence28
- Adoption
- Insufficient
- Hype gap+34
- Incentives38
- Confidence36
A team's Debian 11-to-12 move across 28 hosts put the usual Ansible complaints under load. The failure that actually stopped the run sat in the SSH layer the tool does not own.
Reality
- Evidence34
- Adoption21
- Hype gap+7
- Incentives27
- Confidence40
A dev.to account of retiring Mongo, Redis, Elasticsearch and a vector store makes its strongest argument on consistency rather than cost, and it arrives without a single figure.
Reality
- Evidence24
- Adoption34
- Hype gap+46
- Incentives62
- Confidence30
A side project swapped RabbitMQ, Redis and Elasticsearch for SKIP LOCKED, an unlogged table and a GIN index. The queue arithmetic shows where that choice stops working.
Reality
- Evidence34
- Adoption21
- Hype gap+18
- Incentives38
- Confidence46