What did you do?

When using the "Explain" feature in Prometheus, there's an inconsistency in how millisecond selectors are represented:

• up[1001ms] is explained as up[1s]
• up[1002ms] is explained as up[1s2ms]
• up[1003ms] is explained as up[1s2ms]

What did you expect to see?

Each millisecond selector should consistently and precisely reflect the original query:

• up[1001ms] => up[1s1ms]
• up[1002ms] => up[1s2ms]
• up[1003ms] => up[1s3ms]

What did you see instead? Under which circumstances?

https://prometheus.demo.prometheus.io/query?g0.expr=up%5B1003ms%5D&g0.show_tree=1&g0.tab=explain&g0.range_input=1h&g0.res_type=auto&g0.res_density=medium&g0.display_mode=lines&g0.show_exemplars=0

System information

No response

Prometheus version

prometheus, version 3.2.1 (branch: arythmetics, revision: 63e29dba9a7418d02a4200f60d862e408462380b)
  build user:       reproducible@reproducible
  build date:       19800101-00:00:00
  go version:       go1.24.0
  platform:         linux/amd64
  tags:             netgo,builtinassets,stringlabels

Prometheus configuration file

Alertmanager version

Alertmanager configuration file

Logs