Starting report for a twelve-night restricted protocol: no direct scene manipulation, no summoning, no flying, no testing physics. Every entry treated as an indirect-method exercise only.
The indirect methods I'm using: separation (moving away from the entry point without intending to arrive anywhere), observation (attending to detail without interacting), friction (touching surfaces without trying to change them), sensory confirmation (checking three different sensory modalities in sequence), and passive movement (walking without destination).
Rationale: I've been collapsing states early with direct manipulation. The attempt to change the scene seems to require a kind of cognitive effort that wakes me up. Indirect methods might allow me to stay longer without the cost of that effort.
Night 12 results: seven entries total (three from the device cue, four spontaneous). Average duration 3 minutes 20 seconds. Zero collapses from attempted scene manipulation (because there were none). Two exits from loss of stability unrelated to action.
For comparison: my previous twelve nights with unrestricted methods had five entries, average duration 55 seconds, four collapses directly attributable to manipulation attempts.
The duration improvement is significant. The entry rate is slightly higher, though that may be coincidence over a small sample. Continuing for another twelve nights before drawing conclusions.