compose:debug

Start by reading the full error message and stack trace, then reproduce the failure consistently and check recent changes via git diff. Form a single hypothesis, test it with the smallest possible change, and only implement a fix after confirming the root cause.

Add diagnostic logging at each component boundary, recording what data enters and exits every layer. Run the system once to gather evidence showing exactly which boundary fails, then investigate only that component instead of guessing across the whole stack.

Stop after three failed fix attempts. If each fix reveals a new problem in a different place or requires massive refactoring, the pattern itself is likely flawed and needs architectural review rather than another patch.

Quick fixes address symptoms rather than root causes, so the underlying issue persists and resurfaces elsewhere. Bundling multiple changes at once also makes it impossible to isolate what worked, frequently introducing new defects.

Do not guess. Gather more data through logging and instrumentation until the failure can be triggered reliably, since an unreproducible bug cannot be verified as fixed.