Skip to content
All articles
Engineering19 May 2026·3 min read

Multi-tenant session management: Isolation patterns that actually work

Most systems do not fail because someone chose the wrong database.

Paycux engineering

Most systems do not fail because someone chose the wrong database. They fail because a reasonable decision kept being reasonable long after the conditions that justified it had changed.

This piece walks through how we think about it at Paycux, what we have changed our minds about, and where the sharp edges are.

The problem

That brings us to the problem. Idempotency is not a nice-to-have in any system with retries. Key on an identifier the caller supplies, store the outcome, and return the same answer to the same key rather than doing the work twice.

Make the boundary explicit. When one part of the system can only talk to another through a named interface, you can change either side without a meeting; when it cannot, every change becomes a negotiation.

Pattern 1: Session-per-org

That brings us to pattern 1: session-per-org. Idempotency is not a nice-to-have in any system with retries. Key on an identifier the caller supplies, store the outcome, and return the same answer to the same key rather than doing the work twice.

Measure before you optimise, then measure the thing users feel rather than the thing that is easy to instrument. A p50 that looks fine while the p99 is unusable is a reporting failure, not a performance one.

  • Name the boundary before you cross it
  • Measure what users feel, not what is easy to instrument
  • Make every retried operation idempotent
  • Write down the assumption that would invalidate the design

Pattern 2: Org switching within a single session

Consider pattern 2: org switching within a single session. Idempotency is not a nice-to-have in any system with retries. Key on an identifier the caller supplies, store the outcome, and return the same answer to the same key rather than doing the work twice.

Measure before you optimise, then measure the thing users feel rather than the thing that is easy to instrument. A p50 that looks fine while the p99 is unusable is a reporting failure, not a performance one.

Token scoping rules

That brings us to token scoping rules. Idempotency is not a nice-to-have in any system with retries. Key on an identifier the caller supplies, store the outcome, and return the same answer to the same key rather than doing the work twice.

Make the boundary explicit. When one part of the system can only talk to another through a named interface, you can change either side without a meeting; when it cannot, every change becomes a negotiation.

Make the boundary explicit.

The org claim must be signed, never header-supplied

The org claim must be signed, never header-supplied deserves its own treatment. Make the boundary explicit. When one part of the system can only talk to another through a named interface, you can change either side without a meeting; when it cannot, every change becomes a negotiation.

Measure before you optimise, then measure the thing users feel rather than the thing that is easy to instrument. A p50 that looks fine while the p99 is unusable is a reporting failure, not a performance one.

Refresh token rotation requires reuse detection

Consider refresh token rotation requires reuse detection. Measure before you optimise, then measure the thing users feel rather than the thing that is easy to instrument. A p50 that looks fine while the p99 is unusable is a reporting failure, not a performance one.

Idempotency is not a nice-to-have in any system with retries. Key on an identifier the caller supplies, store the outcome, and return the same answer to the same key rather than doing the work twice.

Regenerate session identifiers at privilege boundaries

Consider regenerate session identifiers at privilege boundaries. Measure before you optimise, then measure the thing users feel rather than the thing that is easy to instrument. A p50 that looks fine while the p99 is unusable is a reporting failure, not a performance one.

Idempotency is not a nice-to-have in any system with retries. Key on an identifier the caller supplies, store the outcome, and return the same answer to the same key rather than doing the work twice.

Where this leaves us

The pattern repeats across every system we have looked at: the hard part is not the mechanism, it is keeping the mechanism honest as the surrounding assumptions change.

If you are working through the same problem and want to compare notes, the docs cover the mechanics and the console shows the behaviour on your own data.

Everything here, already built

Sign-in, enterprise SSO, directory provisioning, roles and an audit trail behind one API. Start with the quickstart and have a working sign-in this afternoon.

Start selling to enterprise customers

Create an account, point sign-in at Paycux, and get back to the part of the product that is actually yours.