The current default signal caching algorithm works like this
1) For given set number and sim number, the code computes fullsky CMB+FG alms.
2) Cache full sky CMB+FG alms
3) Make a copy of full sky CMB+FG alms, and do (patch, array, frequency) specific routines (e.x. beam convolution etc)
4) Cache (patch, array, frequency) specific alms
The potential issue is that the code doesn't distinguish high flux FG and low flux FG when caching CMB+FG alms (step 2 above). So if you generate D56 signal sims with low-flux FG and then try to generate BossN sims with high-flux FG for the same set idx and the sim number, it can load cached CMB+low flux FG alms instead of recomputing alms from scratch.
The current default signal caching algorithm works like this
1) For given set number and sim number, the code computes fullsky CMB+FG alms. 2) Cache full sky CMB+FG alms 3) Make a copy of full sky CMB+FG alms, and do (patch, array, frequency) specific routines (e.x. beam convolution etc) 4) Cache (patch, array, frequency) specific alms
The potential issue is that the code doesn't distinguish high flux FG and low flux FG when caching CMB+FG alms (step 2 above). So if you generate D56 signal sims with low-flux FG and then try to generate BossN sims with high-flux FG for the same set idx and the sim number, it can load cached CMB+low flux FG alms instead of recomputing alms from scratch.