Invariant checker fails on sampled runs, but not on full runs of fleetbench

**Describe the bug**
After merging the fix for #6499, our upsampled runs of fleetbench pass all of the invariant checks. However, for expedience we also use the invariant checker on sampled runs of fleetbench (`-trace_after_instrs 1M -trace_for_instrs 20M -retrace_every_instrs 20M`)

These runs all hit issues in the invariant checker:
```
$/test/drio_build/bin64/drrun -t drcachesim -simulator_type invariant_checker -indir /test/empirical_driver_test/drmemtrace.empirical_driver_test.08501.8356.dir
Trace invariant failure in TTrace invariant failure in T8526 at ref # 9 (0 instrs since timestamp 13348669121513440): An unfiltered thread should have at least 1 instructionTrace invariant failure in T8524 at ref # 9 (
Trace invariant failure in T8508 at ref # 8517 at ref # 9 (9 (00 instrs since timestamp 13348669087323868Aborted (core dumped)
```

```
$ /test/drio_build/bin64/drrun -t drcachesim -simulator_type invariant_checker -indir /test/hot_swissmap_smoketest/drmemtrace.hot_swissmap_smoketest.05085.7339.dir
Trace invariant failure in T5091 at ref # 11337 (1173 instrs since timestamp 13348668791825879): Timestamp does not increase monotonically
```

```
$ /test/drio_build/bin64/drrun -t drcachesim -simulator_type invariant_checker -indir /test/cold_swissmap_smoketest/drmemtrace.cold_swissmap_smoketest.04470.5303.dir
Trace invariant failure in T4476 at ref # 11451 (1292 instrs since timestamp 13348668747489100): Timestamp does not increase monotonically
```

**Expected behaviour**
I would expect the sampled runs of the invariant checker to find a subset of the issues in the full run of the checker. However, I'm not sure what the invariant checker assumes- is it expected to work with sampled runs?

These tests were run with the current HEAD (960d1d7f45b4772cb94f748d0218a9091cbee3bd)