Eight test_disagg_checkpoint_size* tests (08/09/11/13/15/16/18/21) loop on dirty+evict waiting for a ~1% failpoint to roll (__wt_failpoint, src/include/misc_inline.h:281, probability=100). Four then self.fail() - thats WT-18643 - and four skipTest. size13/15 only get 60 rolls, so they skip ~54% of runs and assert nothing about the path they exist to cover.

Use debug_mode.timing_stress_force to force the failpoint and drop the loops. size16 keeps its workload for chain depth, plus one forced hit per phase.