Let (Ω,F,P) be a probability space, let A be a nonempty finite set (the clock labels), let I⊆F be a σ-algebra (the initial data), let R be either a real number with R>0 or +∞, and for each a∈A let Ya=(Yua)u≥0 be a Poisson clock with horizon R on (Ω,F,P) (for R=+∞, a homogeneous Poisson process with rate 1 every path of which is a counting path), such that the family of σ-algebras consisting of I together with the σ-algebras generated σ(Yua:u≥0), one for each a∈A, is independent. Let F⊆F be a σ-algebra (the past), and for each a∈A let Ta:Ω→[0,∞) be an F-measurable random variable (the consumed level of clock a), and assume there is a real number cˉ≥0 with cˉ<R and Ta≤cˉ everywhere for every a. Write R−=R−cˉ, which is +∞ when R=+∞.
For a family c=(ca)a∈A of nonnegative real numbers (caps) let Cc be the event that Ta≤ca for every a∈A, and let Hc be the σ-algebra generated by I together with the variables Yua for 0≤u≤ca and a∈A. Assume the clock-reading bound: for every family of caps c, the event Cc agrees up to an event of probability zero with an event of Hc, and for every F∈F there is an H∈Hc such that the symmetric difference of F∩Cc and H∩Cc is an event of probability zero.
Define the residual clocks
Y^ua=YTa+ua−YTaa(u≥0, a∈A).
Then:
(a) Every path of every residual clock is a counting path, each Y^ua is a random variable on (Ω,F,P), and for every a∈A and all real 0=u0<u1<⋯<up<R− the increments Y^u1a−Y^u0a,…,Y^upa−Y^up−1a are independent, the q-th having the Poisson distribution with parameter uq−uq−1. In particular, when R=+∞, every residual clock is a homogeneous Poisson process with rate 1 on (Ω,F,P).
(b) The family of σ-algebras consisting of F together with the σ-algebras σ(Y^ua:0≤u<R−), one for each a∈A, is independent.